TW201511161A - Systems and methods for automatically verifying correct die removal from film frames - Google Patents

Abstract

A skeleton wafer inspection system includes an expansion table displaceable relative to a camera configured for capturing segmental images of a skeleton wafer on a film frame. During segmental image capture, illumination is directed to the top and/or bottom of the film frame. Segmental images are digitally stitched together to produce a composite image, which can be processed to identify die presence or absence therein at active area die positions having counterpart die positions in a process wafer map. A composite image of a diced wafer on a film frame can also be generated, and used as a navigation aid or guide during die sort operations, or to verify whether a die sort apparatus has correctly detected a reference die prior to die sort operations. A composite image of a skeleton wafer can similarly be generated for use as a navigation aid or guide for film frame repopulation operations.

Description

System and method for automatically checking that a die is properly removed from a film frame

The present invention relates to systems and methods for verifying that the grains carried by the film frame (e.g., provided by wafers adhered to the film frame) have been properly removed from the film frame during the grain sorting operation.

The manufacture of semiconductor components involves a variety of procedures and can generally be divided into front-end processes or back-end processes. The front-end process includes the fabrication of a multi-layer semiconductor device, starting from a bare semiconductor wafer, forming an array of elements or dies on the wafer, wherein each die corresponds to a different semiconductor component (eg, an integrated body that is intended to be incorporated into an integrated circuit package) Circuit chip). After the front-end process, the wafer then undergoes a back-end process, including performing an electrical test of the semiconductor die to determine the electrical or poor quality of the die; and performing a visual inspection of the die in accordance with predetermined test criteria.

Processing wafer map

For each wafer, a "process map" or a "process wafer map" (PW map) is constructed during the initial front-end process. The PW diagram is a digital dataset that records which grains are good and which are defective based on the electrical and visual inspections performed during the front-end and back-end processes.

Wafers are usually circular. The dies located at the edge or at the outermost end of the wafer are typically not used for fabrication. Therefore, the PW map typically stores only the internal regions of the wafer or the "effective region" or "effective grain region" where the die will be fabricated, which is typically less than the total surface area of the wafer. The grains in the effective region may be referred to as effective grains; the grains outside the effective region may be referred to as unprocessed grains, empty shell grains, or mirrored grains due to a "mirror-like" reflection of the unmachined surface. . In most cases, the data set contained in the PW diagram involves only the grains in the effective area of the wafer, plus some information on the empty or mirrored grains to determine and review the position of one or more reference grains. . Therefore, the PW map of the wafer does not fully represent each grid position of the actual wafer, but presents a portion of the entire wafer that substantially or substantially corresponds to the effective area of the wafer, and the die 20 is It is manufactured in this area. Each wafer has a physical wafer identification (ID), such as a bar code, and the PW map of each wafer is digitally linked to the physical wafer ID. The wafer's PW pattern "follows" the wafer during the front-end and back-end processes and is continuously updated as the wafer undergoes each set of electrical and visual inspections. In the PW diagram, the data fields corresponding to the die position of each active area are updated with a specific code that records that the corresponding die 20 is electrically good or bad, and that the grain is good or flawed.

1 shows a partial process block diagram of a typical semiconductor component fabrication system 100 in a prior art, relating to a particular aspect of a backend semiconductor process, as described below.

First visual inspection

Providing a first visual inspection system, apparatus, or module 102 that includes at least one set of image acquisition systems and that is used to perform a first visual inspection of the wafer to find that the wafer is formed in an incorrect manner There are defects, or grains with surface defects or incorrect dimensions.

Bump process and second visual inspection

For various components, after the first visual inspection, the wafer then undergoes a bump process to place or "point" the solder bumps at predetermined locations on the wafer die. The "bump behind" second visual inspection module 104 is used to visually inspect the wafer grains to evaluate the position and flatness of the solder balls on the die, and to find those crystals in which the solder balls do not conform to the predetermined flatness or size. grain.

During the first and second visual inspections, if the die is identified as being visually acceptable, the first and/or second visual inspection modules 102, 104 record one on the PW map in accordance with the position of the die on the wafer. Visually pass the code. For a die having a certain visual acuity, the first and/or second visual inspection modules 102, 104 register a corresponding visual failure code on the PW map.

First part cutting

After the second visual inspection after the bump, the wafer is adhered to the film frame by the adhesive film and sent to the first cutting system, device, or module 106, and the wafer receives the portion or the first time thereon. The dicing, along the upper x - y grid lines, separates the dies on the wafer from one another such that there are gaps between the dies and the dies are thus substantially completely electrically isolated from each other. After this partial dicing, the wafer is not completely cut through and the die is still carried by the lower portion of the wafer.

Electrical test

After partial dicing, the wafer is sent to an electrical test system, device, or module 108 that performs a set of electrical tests on the wafer dies, such as using a wafer probe. For each die, the corresponding electrical test pass or fail code, and the electrical test results of the die are recorded by the electrical test module 108 on the PW map.

Final cutting

After the electrical test, the wafer is sent to the final or second cutting system, device, or module 110 and subjected to a final dicing process whereby the wafer is completely cut through and the individual dies are wafer and Completely separated or isolated from each other. Regarding the second cutting process, the adhesive film adhered by the crystal grains is radially stretched, thereby increasing the separation between the individual crystal grains, and facilitating the selective removal of the crystal grains from the adhesive film frame during the grain sorting operation. This will be further described below. After the second cutting process, the individual dies remain on the film frame, and their relative positions on the wafer before and after the first dicing process, but the intergranular separation is slightly larger. For example, a wafer process may create a "street" of about 40 microns in width between the dies. After the second cutting process, the inter-die separation on the film frame can be up to about 70 - 100 microns, depending on the amount of stretch applied to the film carrying the die.

2A shows a typical fully diced wafer 5 and a corresponding plurality of dicing separation dies 20 adhered to the film frame 12 by a thin layer of viscous film 11. In Fig. 2A, the dicing and separating dies 20 are arranged in a grid which has been laid out on the wafer prior to the dicing process. As shown in FIG. 2A, and as described above, the dicing separation grains 20 are spaced apart from each other by lateral or longitudinal cutting trenches or grid lines 30, 32, which correspond to wafer streets between the dies 20, and crystal The circle is cut along it. Any particular die 20 can be located at a particular grid location on the grids 30,32. The wafer and dicing wafer 5 thus fabricated may include a reference die 21 as will be further described below.

Grain classification operation and third visual inspection

After the second cutting process, the dicing wafer 5 on the film frame 12 is sent to a third, subsequent, or final visual inspection module 112 that performs a third visual inspection procedure to find the dicing separation die 20 Visual inspection caused by cutting operations. During the third visual inspection, the PW map is again updated with the appropriate type of visual pass/fail code.

In this way, after the final/third visual inspection, the PW diagram indicates the cumulative results of all previous electrical and visual inspections, indicating which grains 20 on the film frame 12: (a) good electrical/qualified; b) poor electrical performance; (c) good/qualified visual; and (d) visually flawed. With regard to visual inspection, the PW map can indicate whether each of the dies 20 has one or more specific types of defects, such as size 瑕疵, scratches, chipping, edge unevenness, solder bump coplanarity errors, and/or other types. The error, and whether the flaw falls within the predetermined tolerance standard.

Based on the electrical and visual inspection results on the PW map, the component sorter or die sorting system, apparatus, or module 114 may perform a die sorting procedure to pass the die 20 from the film frame via pick and place equipment or mechanisms. Selective removal, the selective removal operation is based on previous electrical and visual inspection results for each die. Regarding the die classification procedure, the PW map is updated to store the pick and place encoder position, value, or count, typically for the encoder position reference of the reference die 21 (e.g., with reference to the center of the die 21). The encoder position corresponds to the real or actual spatial coordinate of the die 20 in each consideration during the die sorting operation, ie the real/actual spatial position. The pick and place device includes a high resolution imaging system with an image capture device (eg, a camera). The die sorting device 114 includes a film frame expansion station that carries the dicing wafer 5 during the pick and place operation, as will be appreciated by those of ordinary skill in the relevant art.

During grain classification, the poorly performing grains 20 should remain on the film frame 12 so as not to be placed in the downstream or final product. The electrically and visually acceptable dies 20 should be selectively removed from the film frame 12 and sent to a particular destination, typically a tape and reel assembly 120, for use in downstream or final products. In each case, the die 20, which is electrically good but has one or more visual defects (e.g., falling outside of a particular tolerance standard), can be selectively removed from the film frame 12 and sent to one or A plurality of other specific destinations, such as a set of visually unacceptable bins or trays 122, can then be further evaluated or reworked.

The die sorting device 114 selectively sorts the die 20 to a particular final destination, such as a tape and reel component 120 or a particular defective bin or disc 122, which will be specifically finalized according to a set of selectable or programmable classification codes. The die destination is linked to the PW map code corresponding to the electrical test and visual inspection results. In the simplest case considered below, a simplified set of classifications can be defined. As shown in Table 1, only the crystals 20 that are electrically good and visually good, that is, "good grains", are taken out and sent to the volume. The tape assembly 120 (PT); and the poorly-performing crystal grains 20 and the electrically good but poorly-looking crystal grains 20, that is, "bad grains", will not be taken out (NP), so they remain in the glue. On the membrane frame 12.

The dies 20 that should remain on the film frame 12 in accordance with pre-selection criteria may be referred to as "leaved" grains. Since the die 20 is selectively removed, selected, or removed from the film frame 12 during the grain sorting, the dicing wafer 5 has a "hollowed" appearance and becomes a "hollowed wafer." Therefore, before performing the die sorting operation of the dicing wafer 5, the dicing wafer 5 is "not vacant" or "non-hollowed wafer"; during the grading operation of the dicing wafer 5, The more the die 20 is removed, the dicing wafer 5 is gradually turned into a hollow; and after the die sorting operation of the dicing wafer 5 is completed or substantially completed, the film frame 12 carries a hollowed-out wafer.

2B shows a typical hollow wafer 10 corresponding to the dicing wafer 5 of FIG. 2A. The die 20 has been selectively removed from above, leaving only the die 20 remaining thereon: (a) poor electrical, And (b) grains that are electrically good but visually flawed, unacceptable, or unusable, which remain on the hollow wafer 10 as the remaining grains, using the simplified classification code set of Table 1 as a reference. In FIG. 2B, (a) the crystal grains 24 left by the electrical defects are indicated by hatched areas, and (b) the crystal grains 25 which are electrically qualified but visually flawed and should remain on the hollow wafer 10 are And (c) the die 26, which is electrically good and has been removed from the hollow wafer 10, that is, the correctly "empty" hollow wafer die position 26, is represented by a non-shadow or blank area.

Ideally, after the die classification is completed, (a) all of the electrically good dies have been properly removed from the dicing wafer 5; and (b) all of the dies remaining on the vented wafer 10 are It is the die that should be left on the hollow wafer 10. Unfortunately, various errors may occur during grain classification, resulting in grains that are left unintentionally removed from the film frame 12; and/or good grains that should be removed are inadvertently left on the film frame 12. The above-mentioned grain classification errors may have significant adverse economic consequences. For example, if the remaining die is placed in a downstream or final product, such as a packaged IC or board, the resulting product will not be able to meet one or more performance requirements, resulting in a downstream product failure. At this stage, in addition to the cost of manufacturing defective, grain that should be left behind, the economic loss also includes the cost of manufacturing the defective end product. Testing, recalling, and reprocessing the final product may result in significant loss to all parties involved. Therefore, it is important to detect errors during grain classification before the downstream tape is distributed.

Grain classification error categories and how they occur

In general, the causes of grain classification errors can be classified as follows: (A) reference grain detection and re-detection errors; (B) general grain detection failure; (C) grain edge translation error; (D) Other translation errors; and (E) other causes of errors, as detailed below.

(A) Reference die detection and redetection error

In order to accurately perform the grain sorting operation, the pick and place equipment must first correctly identify a reference structure or reference grain. Referring again to Figures 2A and 2B, the reference die 21 corresponds to a particular wafer location relative to the location: (a) the position or coordinates of each of the other dies 20 on the wafer can be referenced or retrieved, thus The dicing wafer 5 can also be referenced or retrieved; and (b) the die-by-die PW map results can be referenced or retrieved such that the dies sorting device 114 can selectively classify the dies 20 into predetermined schedules defined by the classification code. destination. For example, the classification code is simplified according to Table 1. After confirming the position of the reference die 21, the pick and place device can: (a) refer to the good grain on the dicing wafer 5 and the remaining grain position, thus (b) Identifying the dies 20 that should be removed from the dicing wafer 5 during the singulation operation.

In many cases, the reference die 21 contains a unique structural feature on the wafer with a different set of patterns that can be automatically identified by machine vision/image processing algorithms. The unique structural feature can be a different pattern on a particular die, a score, or a particular corner of a plane on the outer edge of the wafer. If the algorithm is robust and powerful, the automatic identification of the reference die 21 is likely to be successful, and the die sorting operation can be performed accurately. However, there are cases where the pick and place device cannot accurately find or identify the reference die 21. This situation can be attributed to one or more factors, such as:

(i) The adhesion of the wafer to the film frame 12 is inaccurate prior to wafer dicing. In this case, the adhesion position of the wafer is slightly deviated from a predetermined position, such as longitudinal and/or lateral +/- 1 mm. Therefore, the camera may not be able to detect the presence of the reference die 21 at the predetermined position, in which case the pick and place operation will stop.

(ii) Film stretching. As described above, the film 11 carrying the dicing wafer 5 is stretched to increase the intergranular separation on the film frame 12 (e.g., from about 40 microns to about 70 - 100 microns) for pick and place. The work is carried out. The displacement error of each of the crystal grains 20 is a net increase in the intergranular separation caused by the stretching. This error will multiply as the dicing wafer 5 is retrieved for multiple die positions. Therefore, the stretching of the film 11 on the film frame 12 may cause the retrieved reference die 21 to slightly deviate from the relative position of the camera, resulting in the pick-and-place device failing to detect the presence of the reference die 21. Once again, the pick and place job will stop.

(iii) The presence of foreign particles on the reference die 21 may cause the imaging system to be unable to detect the presence of the reference die 21 because the foreign particles may distort or change the edge characteristics of the detected die, thus detecting the die The edge detection algorithm used has an adverse effect.

In any of the above cases, since the reference die 21 cannot be detected, the die sorting operation cannot be started. A user or operator intervention is required to teach or retrain the pick and place device identification reference die 21. Through the intervention of the operator, it is possible to select the wrong reference die 21, such as the die 20 adjacent to the originally predetermined reference die 21. If the wrong reference die 21 is selected, a severe systemic error can occur, causing the die sorting device 114 to begin taking the die 20 from the wrong starting point or starting die position on the dicing wafer 5. As expected, the bad, should leave crystal grains may replace the good, the crystal grains to be taken out are taken out, resulting in defects, crystal grains that should be left, that is, "NP" or the unremoved crystal grains on the film frame 12, Send to the tape.

(B) General grain detection error

After identifying the reference die 21, the die sorting device 114 then moves or instructs the docking station to move the dicing wafer 5 to be referenced by die (ie, spanning a predetermined inter-die separation distance), and/or one of the following Proximity grains serve as a reference (when navigating or moving across one or more grid positions without dies 20, "vacancies") to ideally position each of the dicing wafers 5 below the imaging system camera Regional grain 20. At each dicing wafer location where the dies 20 should be left, the imaging system attempts to automatically identify the boundaries or edges of the dies 20. If the imaging system successfully identifies the edge of the die, the imaging system can determine the center point of the die so that any grain can be repositioned relative to the center of the field of view of the camera. Next, the die sorting device 114 can determine that the die 20 under the camera is a good die or a die that should be left according to the grain position indicated by the PW map, so that the die 20 can be selectively taken out or left. On the film frame 12. Based on the determined grain center points, the die sorting device 114 can then move the dicing wafer 5 up to the separation distance between the dies, to the expected location of the adjacent dies 20, and so on.

If the imaging system cannot recognize the edge of the die 20 on the film frame 12 due to the presence of foreign particles or the like on the die 20 (for example, at a lattice position where the die 20 is located but cannot perform automatic edge detection), the die The sorting job will stop. At this point, the operator is required to intervene, reposition or re-retrieve the cut wafer 5 in order to verify the presence of the die and the grain detection under the camera of the imaging system. The operator repositioning/re-retrieving the action of cutting the wafer 5 may result in the erroneous die 20 being placed under the camera, such as the nearest neighbor of the die 20 that should actually be placed under the camera. Therefore, the take-out operation is resumed and continues from the wrong die position, resulting in a systematic take-out error from the die position. This systematic error may cause the die 20 to be removed and sent to the wrong destination (eg, the remaining die is removed and sent to the tape, or a good die is removed and sent to the reject bin), and / Or a good die is left on the hollow wafer 10.

(C) Grain edge translation error

During the die sorting operation, the docking station accepts an instruction to continuously feed each subsequent die 20 to the camera of the imaging system so that each die position can be selectively picked and put. As noted above, imaging systems use machine vision/image processing algorithms to identify grain features, such as grain edges; one or more intra-grain structures or features, such as a row of solder bumps or circuit lines/metallization, The visual or optical properties it has may mimic the optical properties of the grain edges. 3 shows a typical array of dies 20, each of which includes intra-grain features 40, such as one or more rows of solder bumps, which may be misinterpreted as grain edges by an image processing algorithm.

If the image processing algorithm incorrectly identifies a set of intra-grain features 40 as grain edges, the image processing algorithm will incorrectly identify the die center position and subsequently indicate the extension to the next predetermined die. The action of the position will contain a translation error. Depending on the overall grain size and the intra-grain feature location, this grain edge translation error may amount to a significant portion of the overall grain span (eg, about 0.3 mm relative to the center position of the die). In addition, the grain edge translation errors may accumulate through the entire grain array moving or retrieving from the die, as the imaging system may continue to identify the intra-grain features 40 as grain edges incorrectly and unpredictably. In addition, the stretching of the film 11 in which the crystal grains 20 are located may also increase and deteriorate the accumulation of grain edge translation errors.

As shown in Figure 3, this problem typically occurs when there are many of the grains 20 in a column of grains 20 that are left undrawn. In this case, the pick and place device "slights" the die 20 that should be left. When skipped, the imaging system uses an image processing algorithm to identify each die 20 along the column. However, due to the cumulative retrieval error described above, the pick and place device may retrieve the location of the dicing wafer 5 to the erroneous die 20, so the dies that should be left may be incorrectly removed and sent to the tape.

(D) Other translation errors

If it is required to navigate from a certain grid position across a plurality of interlaced grid positions or traverse to a target grid position, the conventional die sorting device 114 uses the following closest die as a reference to navigate from the current grid position to the target grid. Position until you reach the target grid position. The conventional grain sorting device 114 utilizes such navigation techniques to examine the presence and location of a plurality of dies 20 (eg, as many dies 20 as possible) along the navigation path from the current grid position to the target grid location to reduce retrieval or The possibility of locating translation errors accumulates, which may result in grain location errors during the fetch job. It should be noted that the vacant grid position causes automatic detection of grain edges or boundaries within the grid position to verify navigational positioning is not feasible, so the grid position across multiple vacancies increases the likelihood of accumulation of translation errors.

Unfortunately, conventional die sorting devices cannot directly and conveniently navigate from the current grid position across multiple intervening grid locations to the target grid location with high accuracy. In addition, the following one of the closest dies as the reference navigation technology is slow and adversely affects productivity.

(E) Other factors that cause errors in grain classification

Other types of problems can also cause grain classification errors. For example, the PW map used for die categorization is a partial PW map generated from a host PW map present in an individual host system. In some cases, the local PW map may not match the host PW map.

Current grain classification error detection technology

To find that the good die 20 has been properly selected from the dicing wafer 5, the stencil wafer can be inspected manually or via a specific automated procedure, as described in more detail below.

Manually hollowing out wafer inspection

After completing the die sorting operation of the entire wafer of the dicing wafer 10 adhered to the film frame 12, the operator retrieves a hollow wafer 10 from the cartridge. The operator additionally obtains (for example, from a set of back-end manufacturing systems) a printed material that provides a visual representation corresponding to the PW map. This print material has the same size or diameter as the hollow wafer 10 itself, and visually indicates for each original grain position on the hollow wafer 10 whether the die 20 at that location should remain in the hollow wafer 10 on.

The printer overlays the printed data on the hollow wafer 10 under backlight conditions, visually compares the actual wafer 10 and the printed data to manually check whether the poorly-performing crystal grains are incorrectly cut from the hollow. The wafer 10 is removed. Manual inspections are time consuming and error prone, adversely affecting both capacity and output. Depending on the wafer size, die size, and the operator's manual inspection strategy, it may take more than 5 – 20 minutes for an operator to perform a partial inspection of a hollow wafer 10 . More specifically, the printed material may not be perfectly aligned with the hollow wafer 10, and/or the hollow wafer die 20 may not be aligned with the printed material. In addition, the above-mentioned operations are too monotonous, and the possibility of manual error by the operator is large, especially when the technology evolves to gradually increase the diameter of the wafer and the size of the element grain becomes smaller. For example, a die 20 of 2 mm 2 or less is fabricated on a 300 mm wafer, and one wafer carries thousands of die 20 .

In addition, manual inspection can only be performed after the grain classification of the full-box hollow wafer 10 is completed. Therefore, if a systematic grain removal error occurs, affecting a plurality of dicing wafers 5, it is impossible to prevent the systemic singulation from being propagated from one dicing wafer 5 to another dicing wafer 5 in the cartridge. . Therefore, a lot of time, wafer processing resources, and related manufacturing costs are wasted.

Automated hollow wafer inspection

The other current situation is to perform a specific type of automated optical program after the die sorting device 114 selects the die 20 from the dicing wafer 5 according to the wafer PW map and the classification code. In this procedure, the optical inspection system used in the hollow wafer inspection system is the same as that used by the die sorting device 114 during the die sorting operation. The hollow wafer inspection system analyzes the PW map according to the classification code. Based on this analysis, the optical inspection system determines that a limited number of PW regions will be aligned with the corresponding hollowed wafer regions to determine if the die 20 in the hollowed-out wafer region of the consideration has been properly removed from the wafer 10 Remove. Any particular PW pattern corresponds to an array of dies 20 on the wafer. The classification code corresponding to the PW map area indicates which of the dies 20 within the die array are good dies, should have been selected from the vented wafer 10, and which are the remaining dies, which should be left in the stencil 10 on.

The optical inspection system identifies PW regions 1, 2, 3, ..., Z in accordance with predetermined criteria, with respect to the corresponding hollowed-out wafer regions, such as regions having at least a predetermined percentage of undrawn or defective grains, or each The largest percentage of bad grains that can occur in the area. For example, the optical inspection system determines the PW map groups 1, 2, 3, ..., Z having the largest number of defective dies that should remain on the stencil 10, based on the data and classification code groups in the PW map. These Z PW regions are selected to be aligned with the corresponding hollowed wafer regions. The total number of zones Z to be considered and/or the total number of grains D of each zone may be predetermined for each batch of hollow wafers, or may be selected/programmed by the user.

Figure 4 shows a hollow wafer 10 adhered to the film frame 12, and Z = 5 typical hollow wafer areas 18 for comparison with Z = 5 corresponding PW areas. Each zone corresponds to a 5 x 5 array of dies. The hollow wafer 10 shown in FIG. 4 includes some remaining crystal grains 50, and some voided crystal grain positions 52; that is, the crystal grain presence 50 of FIG. 4 is represented by dark shading, and the crystal grains of FIG. Vacancy 52 is represented by a light shade.

The hollow wafer inspection system retrieves the position of the hollowed wafer area in each of the considerations relative to the reference die 21 described above. More specifically, the optical inspection system determines the position relative to the reference die 21 with respect to the position of the reference die 21 with the predetermined grain position as a reference or with the closest die position as a reference, so as to correspond to the position Z = 5 grain locations within the PW pattern, determining the remaining grains 50 remaining in each Z = 5 hollow wafer regions 18, whether they should actually remain at their respective hollow wafer locations And whether the die position 52 that is vacant on the hollow wafer 10 corresponds to the die that should actually be taken out.

Specifically, the optical inspection system (a) selects a first PW pattern for consideration; and (b) along the x and/or y axes, according to a predetermined dicing wafer inter-die separation distance, according to the hollow wafer reference The position of the die moves the hollow wafer 10, thereby positioning a first hollow wafer region 18 corresponding to the first PW pattern within the field of view of the image capture device. The optical inspection system (c) then positions a first predetermined grain position within the first hollowed wafer area 18 below the image capture device, and (d) attempts to determine if the die 50 to be left remains in this Within a predetermined grain position, or whether the first predetermined grain position is a voided grain location 52. The optical inspection system (e) then determines if the PW map indicates, according to the classification code, that the die 20 should remain at the first predetermined grain position, or should be taken from the first predetermined grain position; and (f) if a grain classification error occurs (For example, a defective grain is inadvertently taken out from the first predetermined grain position, or the die 20 is incorrectly left at the first predetermined grain position), and a grain classification error indication corresponding to the first grain position is generated. Next, the optical inspection system (g) is moved to the next or adjacent predetermined grain position in the first hollowed wafer region 18 to determine the presence or absence of grain in the die position and any grain classification errors. Each grain location within the first hollow wafer region 18 is analogous.

After examining each of the die positions in the first hollowed wafer area 18, the optical inspection system (h) moves the hollowed wafer 10 to the second hollowed wafer area 18 according to a predetermined separation distance between the die wafers. And repeating the above-described grain position optical inspection and grain classification error determination for each grain position in the second hollow wafer region 18. This process continues until the Z hollowed wafer areas 18 in question complete the inspection relative to the corresponding Z PW areas. The hollow wafer 10 can then be removed from the die sorting apparatus 114 and the selective die removal during the die sorting operation for the next cut wafer 5 can be considered, followed by the automated hollow wafer inspection process described above.

Unfortunately, the above traditional automated hollow wafer inspection procedures have a number of shortcomings. First, the reference die 21 used in conventional hollow wafer inspection is the same as the reference die used in the die sorting operation, which means that the hollow wafer inspection system may suffer from systematic reference grain position errors, as above. Said. Second, the accuracy of traditional automated hollow wafer inspection depends on the number of wafer areas to be considered. The larger the number of wafer areas examined, the more robust and accurate the evaluation of whether the wrong die 20 is removed. Ideally, each adjacent hollow wafer region 18 spanning the entire hollow wafer 10 should be optically inspected relative to its corresponding PW region to verify that all PW maps and classification codes are identified for removal. The dies 20 have been properly selected from the stencil wafer 10, and all of the dies that are identified as defective by the PW pattern and classification code remain on the stencil wafer 10. Unfortunately, performing inspections of all adjacent hollow wafer regions 18 relative to all adjacent PW regions can be quite time consuming, for example, depending on wafer size and die size, each wafer 10 is between 10 – Between 20 minutes, it has a negative impact on process capacity. Therefore, in order to increase productivity, only a limited number of zones (eg, Z = 5 die zones) are considered in conjunction with an automated hollow wafer sampling algorithm. Unfortunately, if all of the hollowed wafer areas 18 are not inspected relative to the corresponding PW area, it is indicated that there may be a grain classification error that has not been detected. Third, conventional hollow wafer inspection is based on the wafer area and requires a lot of time because images of each die in each selected area need to be taken. Since there are no more than one hundred and hundreds of grains 20 in each zone, the more zones selected, the more time it takes to inspect the wafer. Finally, it may be difficult to accurately locate any particular hollowed wafer area 18 and individual die locations under the image capture device, as it is likely that a large number of dies are vacant from the hollow wafer 10. Therefore, searching for the hollow wafer 10 based on the predetermined inter-die separation distance instead of detecting the grain edge may result in inaccurate positioning of the hollow wafer or difficulty in navigation, which often results in inaccurate inspection results of the hollow wafer.

Therefore, there is a need for a highly accurate and efficient throughput automated system for inspecting the wafer 10 to determine if a grain classification error occurs when the die 20 is selectively removed from the dicing wafer 5 during the die sorting operation.

Manually hollowing out wafer inspections is slow and error prone, as detailed above. Previous automation techniques used high-resolution cameras to perform partial hollow wafer inspections, using a high-resolution camera to examine a limited number of die arrays or regions 18 on the wafer 10, using a high-resolution camera as a reference. The presence or absence of individual grains 20 is present. Previous automation techniques are prone to various errors as described above. In addition, with regard to performing 100% hollow wafer inspection, previous automated methods required a large amount or a large amount of time because high resolution images of each grain position must be taken to verify the effective grain area lattice position on the hollow wafer 10. The vacancy or presence of each die 20 is present. Based on this, although highly desirable, it is not possible to perform a 100% hollow wafer inspection of each of the hollowed wafers 10 after the die sorting operation using the previously automated hollow wafer inspection technique.

It is an object of embodiments of the present invention to overcome all or most of the shortcomings of previously hollowed out wafer inspection systems and techniques. The embodiments described herein provide a hollow wafer inspection procedure that makes inspection operations more robust, efficient, and significantly faster than current hollow wafer inspection techniques, and which are technically simple, inexpensive, and substantially in any die classification Easily performed on device 114, such that 100% of the wafer inspection of each of the hollowed wafers 10 can be performed after the die sorting operation, completely exempting or substantially eliminating systemic or other errors caused by human intervention or other means. .

Particular embodiments of the invention relate to one or more of the following:

(a) Perform the hollow wafer inspection without using the traditional test, inspection and die sorting procedures for the same reference die used prior to the hollow wafer inspection;

(b) Automated 100% hollow wafer inspection, using image capture devices or cameras with lower resolution than image capture devices used in wafer inspection or sorting procedures prior to hollow wafer inspection Large field of view;

(c) Examine the hollow wafer via:

(i) taking a segmented image of a set of hollowed out wafers, using an image capture device having a lower resolution and a larger field of view;

(ii) "splicing" the segmented images together in a digital manner to produce a composite image;

(iii) identifying a reference origin and/or a first/starting grain position within the composite image;

(iii) comparing the data content in the composite image corresponding to the die position of the individual effective region with the data content of the hollow wafer PW map to quickly and accurately determine whether a grain removal error occurs during the classification operation ( For example, if any grains that should be left have been mistakenly removed from the dicing wafer);

(d) generating a composite image of the diced wafer prior to performing a hollow wafer inspection corresponding to the diced wafer, and utilizing the diced wafer composite image as an aid or guide for the pick and place operation;

(e) Using conventional die sorting equipment and incorporating a small number of simple, low-cost components or devices to implement a series or integrated die sorting/hollow wafer inspection system or apparatus that can be used to cut wafers 5 The die sorting operation is performed thereon, thereby producing a hollow wafer 10, and the hollow wafer 10 is automatically inspected immediately according to the corresponding composite image of the hollowed wafer.

Embodiments of the invention may also include the following additional purposes:

(f) Using the diced wafer composite image as a navigation guide for the singularity sorting device/pick and place device, direct navigation to the dicing wafer 5 from which the dies are not removed or one or more target dies on the stenciled wafer 10 Location;

(g) Automatically check if the die classification device has detected the correct reference die 21. Once it is confirmed that the die sorting apparatus has found the correct reference die 21 on the dicing wafer 5 from which the die has not been taken out, the die removing operation on the dicing wafer 5 can be started. Once it is verified that the die sorting device has found the correct reference die 21 on the hollow wafer 10, the recovery operation of the hollowed wafer can be unfolded, during which the die 20 previously removed from the hollowed wafer 10 is placed back into the glue. The film frame 12 has its original effective area lattice position and has a high degree of positioning accuracy.

According to one aspect of the invention, an automated program is provided for generating at least one composite image corresponding to at least one film frame carrying the cutting element, the program comprising: providing a film frame in accordance with a plurality of lattice positions Arranging the cutting element on top, including a set of effective area lattice positions, the component is located inside; taking a set of segmented images of the film frame, each segmented image corresponding to one of the spatial regions spanned by the film frame a predetermined portion, and including at least a set of effective area grid positions; generating a composite image corresponding to the film frame from the segment image group, the composite image including image data indicating (a) each active area of the dwellable element a grid position, and (b) an element presence or component vacancy in each effective area grid position; and processing the composite image via image processing techniques to determine a reference origin and a first effective area grid within the composite image Position, wherein the reference origin may include or be equivalent to a predetermined point of the composite image or a predetermined effective area lattice position, the generation (a) a corresponding predetermined point or predetermined effective area lattice position on the film frame, and (b) a predetermined effective area lattice position on the processed image of the fabricated element, wherein the processed pattern includes or is equivalent to one A data structure that stores data content corresponding to the grid location of each active area, indicating for each valid area grid position that the component resident thereon is pass or fail.

The action of providing the film frame includes at least one of the following: (a) providing a film frame prior to performing a set of component sorting operations and selectively removing the component from the film frame (eg, creating a composite of the film frame prior to removing the component) And (b) providing a film frame after performing the component sorting work group (eg, producing a composite image of the film frame after removing the component).

The act of capturing the segmented image set includes directing the illumination toward at least one of: (a) a top surface of the film frame, with the first set of illumination parameters as a reference, and (b) a bottom surface of the first film frame, The second set of lighting parameters is used as a reference. The first set of illumination parameters and the second set of illumination parameters facilitate reliable image processing, identifying the presence of elements in the grid position, the vacancies of the elements from the grid positions, and the grid lines defining each grid position. The action of capturing a segmented image group also includes capturing each segmented image, the image capturing device used having a lower image capturing device than the microscopic or sub-micron level used to inspect individual components, respectively. Resolution and a large field of view (for example, several times larger). The segmented image set may include a plurality of images corresponding to predetermined portions of the entire area of the film frame, and generating the composite image includes splicing the individual segment images within the segmented image group in a digital manner. The splicing action may involve connecting individual segmented images at least partially corresponding to adjacent regions of the film frame by aligning the following individual segmented images: (a) overlapping along a common boundary set, corresponding to component edges of the same component and / or component features, and / or (b) directly adjacent to each other, with the encoder value as a reference, which indicates the relative actual space coordinates of the film frame during the segmentation of the segment image group. The composite image can be an image of the entire wafer.

A prescription can be evaluated before the composite image is produced, the prescription including at least the following items: (a) grid data, including some horizontal grid lines and some longitudinal grid lines, which define a plurality of grid positions; (b) Image capture device parameters, including image capture device resolution relative to component size and wafer size, and image capture device field of view; (c) at least one set of illumination parameters used to control incidence during segmentation of the image group Lighting characteristics on the film frame; (d) some segmented images to be photographed; (e) degree of overlap between individual segmented images within the segmented image group; (f) a set of reference origins a parameter indicating or defining a reference origin position relative to a plurality of lattice positions; (g) a first effective region lattice position indicating a first component position within the effective region lattice position group; and (h) a A set of parameters that can be used to examine the position of the reference grain detected by the die sorting device.

The reference origin is different from the reference grain position on the film frame used to perform the component classification work group. The act of processing the composite image includes identifying at least a portion of the plurality of ruled lines in the composite image corresponding to at least a portion of the plurality of cutting grooves between the elements on the film frame; The prime region defines a plurality of lattice positions, the boundary of which is surrounded by adjacent horizontal grid lines and adjacent vertical grid lines in the plurality of grid lines; and the reference origin and the reference origin and the first effective region grid position are utilized The predetermined spatial relationship between the plurality of lattice positions determines the effective area grid position.

The act of processing the composite image may further comprise analyzing pixel data within the composite image of each effective region lattice location to determine a corresponding effective region lattice location or component vacancy of the component on the film frame; and for the effective region lattice Each valid area grid position in the location group compares the data content of the composite image with the data content of the processed image, for example, for hollowing out wafer inspection. The act of aligning the data content includes comparing the pixel values corresponding to the grid positions of each of the effective regions in the composite image with the digit codes corresponding to the grid positions of the effective regions on the processed image. The program may include automatically determining: (a) whether one or more components were incorrectly removed from the film frame during the component sorting operation, and/or (b) whether one or more components were incorrect after the component sorting operation The ground is left on the film frame.

The composite image produced for the film frame prior to the execution of the component classification work group can be used as a navigation aid or guide during a set of pick and place operations to locate at least one of the following: (a) in the same film frame or The effective lattice target grid position relative to the pick-and-place equipment during or after the actuator sorting operation group on the different film frames; (b) during the component sorting operation group on the same film frame or different film frames, as opposed to Each valid area grid position of the pick and place device; and (c) at least some effective area grid positions within the plurality of grid positions relative to the pick and place apparatus during a set of film frame restoration operations, the restoration object is classified by the component operation Groups of components that are removed from the same or different film frames.

The film frame may be a first film frame carrying a first dicing wafer in a plurality of dicing wafers, wherein the composite image of the film frame comprises a composite image of the first dicing wafer, and wherein the composite image is used Actions such as navigation aids or guides include: storing a composite image of the first cut wafer as a batch navigation composite image that can be used as a navigation aid or guide for each wafer in the batch; selecting a second glue a film frame carrying the second dicing wafer in the batch; generating a composite image of the second dicing wafer; using image registration techniques to determine and correct the composite image of the batch navigation and the second dicing wafer The rotation between the images is offset.

The program can include identifying a current die location on the second dicing wafer and a corresponding current dies location in the composite image of the second diced wafer; identifying a target dies location and batch of the second diced wafer Navigating the corresponding target grain position in the composite image; using the batch navigation image and the image space to the actual space conversion coefficient, calculating the relative encoder position corresponding to the current die position and the relative position corresponding to the target grain position Encoder position; generating an updated encoder position based on each calculated relative encoder position; and using the updated encoder position to navigate the pick and place device directly from the current die position to a set of boundaries of the target die position Inside.

The program can include verifying that the die sorting device detects the correct reference die on the second dicing wafer by retrieving a relative encoder position offset corresponding to the auxiliary reference dies from the recipe; classifying the dies The device directly navigates from the candidate reference die position or the candidate reference origin to a predetermined position of the auxiliary reference die; automatically determines whether the auxiliary reference die appears at a predetermined position of the auxiliary reference die; and according to whether the auxiliary reference die appears in the auxiliary Referring to the predetermined position of the die, it is checked whether the candidate reference die is the correct reference die.

The process can include selectively removing the die from the second dicing wafer to form a stenciled wafer; generating a composite image of the stenciled wafer; and utilizing image registration techniques to determine and correct the batch navigation composite image Rotational offset between the composite image with the hollowed out wafer.

The program can include retrieving a relative encoder position offset corresponding to the auxiliary reference die from the recipe; navigating the pick and place device directly from the candidate reference die position or candidate reference origin on the hollowed wafer to the auxiliary reference die a predetermined position; automatically determining whether the auxiliary reference die appears at a predetermined position of the auxiliary reference die; and determining whether the candidate reference die is the correct reference die according to whether the auxiliary reference die appears at a predetermined position of the auxiliary reference die; Whether the candidate reference die is the result of the inspection of the correct reference die, starting a set of film frame restoration operations, the film frame recovery operation group includes returning the die that has been removed from the second cutting wafer back to the second film The frame is used to form a substitute second dicing wafer, a hollow wafer composite image and a batch navigation composite image are used as a film frame recovery operation group, and navigate to different crystals on the second film frame Auxiliary or guide to the position of the grain.

Through the parallel computing program, the generation of the composite image can occur simultaneously with the execution of the component classification job group.

According to one aspect of the invention, a system is provided for generating at least one composite image corresponding to at least one film frame carrying a cutting element, the cutting element being located in an effective lattice of a plurality of lattice positions In the area, the system includes: a processing unit; an extension or wafer table for carrying and fixing the film frame; and a set of illumination sources used during the loading of the film frame by the extension or wafer table. To guide the illumination toward the film frame; an image capture device; and a memory that stores at least a set of program instructions that, when executed, cause the processing unit to execute at least a portion of the program.

The processing unit is coupled to a component sorting device for performing a set of component sorting operations on the film frame that involves selectively removing components from the film frame in accordance with the contents of the processed image. The system can be (a) integrated into a component sorting device for performing component sorting operations, or (b) separated from component sorting equipment.

The illumination source set includes at least one of: (a) a first illumination source group for directing illumination toward the top surface of the film frame in accordance with the first illumination parameter set during the segmentation of the segment image group, and (b) And a second illumination source group for guiding the illumination toward the bottom end surface of the first film frame in accordance with the second illumination parameter set during the shooting of the segment image group. The image capture device has a lower resolution and a larger field of view than an image capture device used to inspect micro- or sub-micron 瑕疵 of individual components.

The memory stores a prescription that includes at least the following items: (a) grid data, including some horizontal grid lines and some longitudinal grid lines, which define a plurality of grid positions; (b) image capture device parameters, Include image resolution device resolution and image capture device field of view relative to component size and wafer size; (c) at least one set of illumination parameters used to control incidence on the film frame during segmentation of the image group Lighting characteristics; (d) some segmented images to be photographed; (e) degree of overlap between individual segmented images within a segmented image group; (f) a set of reference origin parameters, which indicate or Defining a reference origin position relative to a plurality of grid positions; (g) a first effective area grid position indicating a first element position within the effective area grid position group; and (h) a set of parameters, Check the position of the reference grain detected by the die sorting device.

In the present invention, the specific elements in the specific drawings are described, or the specific element numbers in the specific figures are considered or used in the corresponding description materials, and may be included in other drawings or related description materials. The same or similar component or component number is indicated. Use "/" for "and/or" unless otherwise indicated. As used herein, the term "group" corresponds to or is defined as a finite non-empty set of elements that mathematically exhibits a cardinality of at least one (ie, a set referred to herein may correspond to a single element group or multiple elements). Group), as defined by known quantities (for example, Introduction to Mathematical Reasoning: Numbers, Sets, and Functions "Chapter 11: Characteristics of Finite Sets" (as described on page 140) by Peter J. Eccles, Cambridge University Press (1998)). In general, one element of a group may include or be equivalent to a system, device, device, structure, structural feature, purpose, program, actual parameter, or value, depending on the type of collection in the context. The specific numerical values or ranges of values recited herein are understood to include or are equivalent to the approximate value or range of values, such as +/- 20%, +/- 15%, +/- 10%, or +/- of the stated value. Within 5%.

Hollow wafer inspection in accordance with an embodiment of the present invention

In accordance with an embodiment of the present invention, the core of the hollow wafer inspection involves taking a segment or partial image of the hollow wafer 10 after the grain classification. The sum of all segmented images, when properly aligned, constitutes a complete image of the hollowed out wafer. Each segmented image captures many of the dies in the segment at a time, and the image capture device used has a lower resolution and a larger field of view than the image capture device used in the die sorting operation. After the segmented images have been taken, they are "spliced" in a digital manner (e.g., using an image processing algorithm) to form a composite image of the entire hollow wafer 10. The composite image is analyzed in a digital manner to determine whether the die 20 is present on the hollow wafer 10 or vacant from the hollow wafer 10 for each hollow wafer grain position corresponding to the effective grain area of the wafer. More specifically, an image processing algorithm is utilized to determine the presence or grain vacancies of the grain locations of each active region within the composite hollow wafer image, which corresponds to each corresponding on the actual hollow wafer 10 Effective area grain position. According to various embodiments of the present invention, the die presence or the die vacancy data is further coded for each composite image effective region die position of the hollow wafer selection map, for example, the values "1" and "0", respectively. Express it. The hollow wafer selection map is digitally compared with the data generated by the PW or PW map to determine if the dicing wafer 5 has been correctly classified. The above procedure will be explained in more detail below.

Typical hollow wafer inspection system configuration

Depending on the details of the embodiment, there may be one or more portions of the automated hollow wafer inspection system 200: (a) becoming a series or related portion of the component sorter/die sorting device or die sorting and inspection system 114; (b) becoming a centralized system or a hollow wafer inspection center that can receive the hollow wafer 10 from a set of component sorter/die sorting devices 114; or (c) become a stand-alone system.

Typical tandem die classification and hollow wafer inspection system

5A shows a block diagram of a portion of a process corresponding to the exemplary semiconductor component fabrication system 101 of FIG. 1, but including an automated hollow inspection system (SIS) or a hollow wafer inspection system (SWIS) 200 in accordance with an embodiment of the present invention. In various embodiments, the SWIS 200 is integrated into or integrated into a component sorter/die sorter or component/die sorting and inspection system for performing tandem hollow wafer inspection 115, which is The conventional component sorter/die sorting device/die sorting and inspection system 114 is integrated, such as a system using iSORT (Singapore Semiconductor Technology and Instruments (STI) Private Limited). The SWIS 200 can transmit specific hollow wafer inspection results to the tape and reel assembly 120 or other systems/equipment to take remedial action after the hollow wafer inspection.

Figure 5B is a generalized block diagram of a die sorting and inspection system for use in a tandem hollow wafer inspection 115 in accordance with an embodiment of the present invention. The die sorting and inspection system 115 includes a film frame cassette 150 that can be removed from the cassette and transported to the film frame 12 carrying the dicing wafer 5 using an automated transfer mechanism (e.g., a robotic arm, not shown). A film frame retention/expansion station 210 that is used to expand the film 11 of the film frame prior to the die sorting operation and is used to hold the film frame 12 during grain sorting and hollowing wafer inspection operations, It will be further described. The die sorting and inspection system 115 further includes a pick and place apparatus 160 for selectively removing or picking the die 20 from the dicing wafer 5 and delivering it to a predetermined destination. The stage 210 can be moved as xy, z, and Q axes (ie, axis or θ) and is typically displaced at least two dimensions (eg, along the x-y axis) and rotated about the Q axis to present the cut in the correct orientation. Wafer 5 for picking up operations. The extension table 210 can also be moved in a Z-axis to stretch or expand the film 11 on the film frame 12 for picking up the die from the film frame 12 during pick and place operations.

The die classification and visual inspection system 115 also includes a first camera 170 for capturing a high resolution image of each of the extracted dies 20; a second camera 172 for performing each of the extracted dies 20 a top inspection and a bottom inspection; a third camera 174 for performing a 4-side inspection of each of the removed dies 20; a fourth camera 180 for performing a tape-sealed bag inspection; and a A five camera 182 is used to perform device inspection in the sealed bag.

The die classification and visual inspection system 115 further includes a segmented image capturing device 220 (hereinafter referred to as "SWIS image capturing device" or "image capturing device"), such as a camera or a digital image sensor, which can be expanded The stage 210 is positioned below it to capture a segmented or regional image of the surface area of the hollowed wafer, as further described below. The die classification and visual inspection system 115 additionally includes a set of illumination sources/elements associated with the image capture device 220, as will be described in more detail below.

In accordance with a particular embodiment of the present invention, a conventional type of die sorting and inspection system/die sorting device 114 can be used for tandem hollow wafer inspection, the structure and function of which are described in the published patent application WO2009128790, US20100232915, and SG 201103425-3. The structural and functional aspects of such conventional grain classification and inspection system/die sorting apparatus 114 will be immediately apparent to those of ordinary skill in the relevant art.

Other representative SWIS configurations

Other semiconductor manufacturing system configurations involving one or more sets of Hollow Wafer Inspection System (SWIS) 200 may be provided, such as a configuration in which the SWIS 200 is separate from the component sorter/die sorting device 114. For example, Figure 5C shows a block diagram of portions of a typical semiconductor fabrication system associated with die classification, including a plurality of die sorting devices 114 for delivering hollowed wafers to a common or centralized portion in accordance with an embodiment of the present invention. The wafer inspection system 200 is hollowed out.

According to various embodiments of the present invention, a complete hollow wafer inspection can be automatically performed after the die sorting operation, corresponding to inspection or analysis of the actual die selection from the hollow wafer 10, and inspection of the effective grain area of the wafer 100% of the grain 20 may be present or previously may be present in a hollow wafer location or grid location. In this manner, in accordance with an embodiment of the present invention, a set of SWIS 200 can automatically: (a) perform a 100% hollow wafer inspection/analysis whereby the actual crystals at each grain location on the hollow wafer 10 can be automatically verified. The grain presence 50 and the actual grain vacancy 52; and (b) the PW pattern relative to the diced wafer (or the PW pattern of the stencil wafer), and any associated classification code sets are considered to determine if there are any grain classification errors.

The face of a typical SWIS component

Figure 5D is a schematic illustration of a portion of a program of a hollow wafer inspection system (SWIS) 200 in accordance with an embodiment of the present invention. As described above, the SWIS 200 can be in series or internal to the die sorting device 114, or can be a hub or stand-alone system.

SWIS film frame retention / extension table

In one embodiment, the Hollow Wafer Inspection System (SWIS) 200 includes an extension station 210 for stably holding a film frame 12 supported along the periphery or edge of the film frame 12. More specifically, the extension table 210 includes a set of surrounding support members 212 for receiving and stabilizing the surrounding portion of the retention film frame 12 (e.g., via a slot structure) via the adhesive film 11 on which the die 20 is located, The hollow wafer 10 carried by the film frame 12 is positioned at a predetermined orientation and position above a notch or opening 215 in the extension table 210. The extension table 210 also includes a bracket member 214. The surrounding support member set 212 is longitudinally displaceable relative to the bracket member to radially stretch the film 11 of the cut wafer 5 to increase the separation between the crystal grains 20 so as to be crystallized. Remove the pellets. The extension station 210 may correspond to or be equivalent to a conventional type of film frame retention, expansion, and displacement device that occurs in conventional grain sorting apparatus 114 in a manner that is immediately apparent to those of ordinary skill in the relevant art. When the SWIS 200 is in series with the die sorting device 114, they can share a common platform, and the SWIS 200 can use the same expansion station 210 as the die sorting device 114 (eg, the same expansion station 210 can be used for die sorting and hollowing out wafers) check homework).

SWIS image capturing device and illumination source

To facilitate hollowing out wafer inspection operations, the SWIS 200 also includes (a) an image capture device 220 (eg, a camera or digital image sensor) that can be positioned below the extension station 210 for capturing hollow wafers. a segmented area image of the surface area; and (b) one or more illumination sources/elements 230, 232 for illuminating the hollowed out wafer 10 such that the composite image of the hollowed wafer is generated from the corresponding segmented image , the presence of crystal grains, grain vacancies, and grid lines 30, 32 can be immediately recognized. The illumination source 230, 232 includes at least a first or lower/rear illumination source set 230 for directing illumination from beneath the film frame 12 toward the portion of the hollow wafer 10 being imaged by the upper image capture device 220; A second or upper/top illumination source set 232 is used to direct illumination toward the upper side surface of the hollow wafer 10. In some embodiments, the SWIS 200 can additionally configure a set of optical elements disposed between the extension station 210 and the image capture device 220 for image capture operations.

The first illumination source set 230 is used to illuminate an area that is at least as large as the size of each segmented image, typically illuminating a slightly larger area. Accordingly, the first illumination source set 230 is used to illuminate an area that is at least as large as the field of view of the SWIS image capture device 220. Depending on the details of the embodiment, the first illumination source set 230 can include one or more types of light-emitting elements, such as an array of LED lights. The second illumination source set 232 is used to direct the illumination toward the upper or top surface of the film frame 12 at an appropriate angle (angle adjustable/optional) such that the light output of the second illumination source set 232 can be removed from the wafer 10 The surface is reflected to the image capturing device 220. In some embodiments, the second illumination source set 232 can include or be equivalent to a surrounding light source, such as an annular illumination device (eg, including one or more columns of circumferentially disposed LED lights) disposed on the SWIS image capture device 220. Between the extension station 210 (eg, within an illumination gap between the image capture device 220 and the extension station 210).

The illumination sources 230, 232 output light in accordance with a particular illumination parameter, such as a set of illumination intensities, a set of illumination wavelengths, and/or a set of angles of incidence to accurately and reliably automatically identify pixel regions within the captured image, corresponding to (a) Occupied hollow wafer grid position/left grain; (b) voided wafer grid position/vacant grain area; and (c) cutting trench or grid lines 30, 32. In various embodiments, the angle of illumination, illumination intensity, and/or illumination wavelength incident on the adhesive film 11 of the film frame can be altered, adjusted, or selected to facilitate automated identification. In some embodiments, the SWIS inspection prescription may include or specify (eg, in an alternative or programmable manner) a particular illumination source 230, 232 or a desired segment and specific for each active illumination source 230, 232 Lighting parameters.

SWIS image capture device field of view

The SWIS image capturing device 220 has a field of view sufficient to capture a segmented image of the hollowed-out wafer space region, that is, an image of a segment portion or region of the entire surface area of the hollowed out wafer, wherein any particular segmented image is typically taken A significant or non-trivial segment of the overall surface area of the wafer is hollowed out (eg, as compared to the area of the individual dies 20). For example, the SWIS image capture device provides a field of view that is at least slightly larger than 1/36, 1/25, 1/16, 1/9, 1/6 of the surface area of the bezel surrounding the overall cross-sectional area or surface area of the hollowed out wafer. 1/3, 1/4 or 1/2, so that each segmented image is captured at least slightly more than 1/36 of the surface area of the hollowed out wafer, 1/25, 1/16, 1/9, 1/6, 1 /3, 1/4 or 1/2. In addition, the SWIS image capture device 220 has a resolution sufficient to facilitate identification of the dicing trenches or grid lines 30, 32 and to facilitate determining that the dies 20 are present at or corresponding to a particular hollow wafer dies/lattice location. 30, 32 positions, or vacancies.

In an exemplary embodiment, the SWIS image capture device 220 provides: (a) a relatively large, significantly larger image capture device than is used to visually inspect or identify/find individual die 20 or small array die 20 Or a much larger field of view; and (b) a smaller, significantly smaller, or much smaller resolution than an image capture device used to visually inspect or identify/find individual die 20 or small array die 20 degree. As is known to those of ordinary skill in the relevant art, the image capture device used to visually inspect individual die 20 or small array die 20 (e.g., during the first, second, or third visual inspection described above) is High or very high resolution images are captured in the field of view to determine if very small or very small (eg, micron or even sub-micron) crystals are present on the integrated circuit structure in the spatial extent of any single die 20. Intragranular optical enthalpy.

In contrast, in accordance with an embodiment of the present invention, SWIS image capture device 220 allows for the capture of a set of segmented images, wherein each segmented image typically captures a plurality of dies 20 (eg, a plurality of dies 20), and The corresponding surface area of the wafer is significantly larger than that of the first, second, or third visual inspection operation or the single crystal 20 or small array of crystal grains for grain identification/positioning during the grain sorting operation. 20 surface area. The SWIS 200 produces a composite image of the entire hollow wafer 10 from the segmented image. Segmented images and composite images provide sufficient resolution to allow SWIS 200 to: (a) detect or reconstruct segmented images and grid lines within the composite image separately; and (b) evaluate the presence of a composite image of the grains Like the grid position or the grain vacancy, it corresponds to the hollow wafer grid position defined by the actual grid lines 30, 32 of the hollowed out wafer.

In SWIS operations, it is not necessary to identify the presence or absence of defects in the grains (eg, micron or sub-micron optical defects). Therefore, the SWIS image capturing device 220 will have: (i) a lower resolution than the die-dividing camera, and (ii) a relatively large field of view, allowing each segmented image to capture the surface area of the hollowed out wafer. Heavy or fragment. This means that a significantly smaller image can be taken to produce a composite image of the entire processed hollow wafer 10 than the camera used by the die sorter to take an image of each die 20 over the entire surface area of the hollowed wafer. Like, including its grid line 30, 32. For example, depending on the details of the embodiment, for small or very small sized molds 20 (eg, 2 mm x 2 mm or smaller) fabricated on large wafers (eg, 300 mm wafers), SWIS The image capturing device 220 has a field of view that allows ten, tens, hundreds, or 1,000 dies 20 to be displayed on each segmented image taken. Further, the SWIS image capturing device 220 need not be a color image capturing device, which may be a black-and-white or monochrome image capturing device.

In view of the above, in accordance with embodiments of the present invention, hollow wafer inspection does not involve capturing many high resolution images of individual dies 20 (eg, when the effective grain area of a wafer includes hundreds or thousands of When the grain is 20, hundreds or thousands of individual grain images). This saves a significant amount of time or a significant amount of time compared to previously hollowed out wafer inspection techniques (especially for 100% open wafer inspections). Moreover, since various embodiments of the present invention may use a lower cost low resolution image capture device 220, and several embodiments use the same extension station 210 as the die sorting device 114, the hollowed out crystal in accordance with an embodiment of the present invention The round inspection system is cheaper to implement than previously hollowed out wafer inspection systems.

In a typical embodiment, the hollow wafer inspection operation is performed on an 8-inch wafer having a grain size of not less than 1 x 1 mm2, and the SWIS image capturing device 220 can be used to capture a set of nine segment images. It spans the entire surface area of the hollowed-out wafer, has a field of view of approximately 80mm x 80mm, and has a digital image sensor that can capture a 4 million pixel digital image, providing a resolution of approximately 5μm. In the exemplary embodiment, the SWIS 200 can inspect the 100% effective grain area of the 8 hollow wafers and verify that the grains are properly removed within 5 minutes, within 3 minutes, or within approximately 2 - 2.5 minutes.

One of ordinary skill in the relevant art will appreciate that if the grain size is smaller, the resolution of the SWIS image capture device 220 can be increased and relatively more segmented images can be taken. With regard to additional/other typical SWIS embodiments, Table 1 below provides typical SWIS image capture device field of view and resolution parameters for specific grain sizes and wafer sizes, as well as for imaging the entire surface area of the wafer 10 The number of segmented images.

Easy to hollow out the inspection of wafer inspection

As described above, the SWIS 200 provides illumination sources 230, 232 that facilitate the reliable identification of the following items using image processing: (a) hollowing the wafer cutting trenches or grid lines 30, 32, which define the lattice locations. During the die sorting operation, depending on the electrical test results, the visual inspection results, and the classification code, the die 20 may remain in the grid position on the dicing wafer 5, and may have been taken out and classified; (b) The grain has a specific hollow wafer grid position; and (c) the grain is vacant from other hollow wafer grid locations.

As will be appreciated by those of ordinary skill in the relevant art, during dicing of the wafer to separate each of the dies 20 above, the saw cuts (not cut through) the film 11 on the film frame 12 as it cuts through the wafer. As shown in FIGS. 6A and 6B, the cutting process forms a partial depth channel 34 on the film 11. The channel 34 is located at the bottom of each of the dicing wafer 5 and the dicing trenches 10 or the bottom lines 30, 32 of the hollow wafer 10.

The remaining grains, the vacant grain locations, and the dicing trenches 30, 32 will affect the illumination that is directed to the hollow wafer 10 in different ways. More specifically, the illumination directed to the hollow wafer 10 is reached at any particular point within the image capture plane corresponding to the SWIS image capture device 220 (eg, the image capture plane in which the image sensor is located). The extent of this point depends on: (a) whether illumination is provided from above and/or below the hollow wafer 10; and (b) the remaining grain, the location of the vacant die, or whether the cutting trenches 30, 32 are along The optical path of the illumination is presented. Thus, depending on the lighting conditions of the hollowed out wafer, the remaining crystal grains, the location of the vacant grains, and the imaging of the dicing trenches 30, 32 will each have different optical or visual characteristics.

Regarding the illumination provided from below the hollow wafer 10, the remaining dies will block illumination from passing to the image capture device 200; the vacant die position allows illumination to pass through the film 11 towards the image capture device 220; A portion of the depth channel 34, the hollowed-wafer grids 30, 32 will at least partially disperse (possibly disperse) the illumination that travels beneath the hollow wafer 10 toward the image capture device 220. Even if dispersed, the image capture device 220 will still capture some of the illumination of the passing lines 30, 32. Regarding the illumination provided above the hollowed wafer 10, the remaining dies will reflect the top illumination, unlike the vacant die position and grid lines 30,32.

7A shows a composite image 1000 of a hollow wafer 10 under typical illumination conditions, providing illumination to the top and bottom ends of the hollow wafer 10 during the capture of the segmented image, such as via a first or lower side illumination source 230 And a second or upper illumination source 232. Figure 7B shows an enlarged portion 1002 of the hollowed-wafer composite image 1000, which corresponds to a 2 x 7 array of hollowed-wafer grid locations, defined by lateral and longitudinal grid lines 30, 32.

In the illumination conditions of Figures 7A and 7B, the remaining grains 50 exhibit dark or very dark shaded areas; the voided grain locations 52 exhibit a shadowless or minimal/lightly shaded area; and the cut trenches 30, 32 A relatively dark or medium dark narrow line is then defined which defines the lattice position of the hollowed out wafer. In general, within a composite image 1000 produced by segmented images taken under appropriate lighting conditions: (a) the remaining dies 50 correspond to a pixel array having a first average intensity; (b) The vacant grain position 52 corresponds to a pixel array having a second average intensity that is different or significantly different from the first average intensity; and (c) the dicing grooves 30, 32 correspond to a narrower having a third average intensity A pixel line that is different from the second average intensity and that may be different or significantly different from the first average intensity. The remaining grains 50, the vacant grain locations 52, and the imaging of the dicing trenches 30, 32, the relative optical or visual properties depend on the particular conditions under which the top and/or bottom ends of the hollow wafer 10 are illuminated (eg, , which illumination source 230, 232 actuation, and illumination parameters such as illumination intensity, wavelength, angle of incidence, etc. of each illumination source 230, 232), and grain size and grain defined by the cut trench/grid 30, 32 Separation (depending on the width of the ruled line and the extent to which the film 11 is stretched).

In various circumstances, specific illumination conditions, including actuation or inactivity, may be established, adjusted, or modified to increase the overall imaging contrast of the grid lines 30, 32 relative to the remaining grain 50 and/or the voided grain location 52. The particular illumination source 230, 232 and the illumination parameters used to select the illumination used, such that the imaging of the grid lines 30, 32, the remaining dies 50, and/or the vacant die position 52 are more easily imaged by the image processing algorithm Or visual difference.

Moreover, in many cases, structural features, such as leaving the metallization or solder bumps on the upper side surface of the die 50, will be different from (eg, stronger) vacant grain locations 50 and grid lines 30, 32 The way to reflect the top illumination. Thus, the presence of such structural features on the remaining grains 50 enhances the likelihood that the image processing algorithm will reliably distinguish between the grain 50 and the vacant grain position 52 and the grid lines 30, 32. For example, Figure 7D shows a partial segmented image of the hollowed wafer 10 taken under typical top lighting conditions. As shown in Fig. 7D, under this illumination condition, the grains 50 remaining in the segmented image are easily optically or visually distinguished from the vacant grain locations 52 and grid lines 30, 32.

In various embodiments, proper hollowing wafer illumination conditions during the segmentation of the image may be balanced against the correct contrast between the gridlines 30, 32 relative to the remaining die 50 and the voided die location 52. One of ordinary skill in the relevant art will recognize that if the illumination intensity provided by the first illumination source set 230 to the underside of the hollow wafer 10 is too great, the contrast between the gridlines 30, 32 and the voided die location 52 will be low. . However, in many cases, the underside illumination should not be too low, otherwise the grid lines 30, 32 between the remaining dies 50, 52 are illuminated and photographed, which may be detrimental to the reliable discrimination of the grid lines by image processing. 30, 32 with the remaining grain 50 and/or the vacant die position 52. In some embodiments, collimated light (e.g., collimated white light) can be used to increase the likelihood of distinguishing the grid lines 30, 32, the remaining grains 50, and/or the voided grain locations 52 from imaging and image processing. Sex. It should be noted that more than one (at least one) of the segmented images of each segment can be taken under different illumination for image analysis or comparison.

As described above, the first illumination source set 230 is an area for illuminating at least as large as the field of view of the SWIS image capture device. In various embodiments, the second illumination source set 232 is used to illuminate the entire upper side surface of the hollow wafer 10; however, in some embodiments, the second illumination source set 232 illuminates less than the hollow wafer 10 The entire surface. To provide an appropriate or ideal contrast in each segmented image to distinguish between grids 30, 32 and detect grain presence 50 and die vacancies 52, create or select a set of segments before taking a segmented image Appropriate lighting types and lighting parameters (eg, forming part of a hollow wafer inspection prescription). Such types and parameters of illumination may depend on the type and size of the dies 20 that are fabricated from a batch of wafers under consideration.

Processing unit and memory

The SWIS 200 further includes a processing unit 250 and a memory 260 that is coupled to the SWIS image capture device 220. The processing unit 250 can execute program instructions (e.g., software) stored in the memory 260, including one or more sets of program instructions for controlling or executing the automated hollow wafer inspection and/or analysis program of embodiments of the present invention. For example, the memory 260 can include a hollow wafer inspection module 262 that includes the above-described program instruction set.

The memory 260 can further include at least the following items:

I. A hollow wafer inspection configuration/check prescription memory 264 that can store or reference any particular wafer in question (eg, in a batch of wafers), hollow out wafer inspection configuration/setting data, or hollow wafer inspection The prescription will be further described below;

II. An image memory 266 for storing a set of segmented images taken corresponding to the area, portion, or segment area of the hollow wafer 10, and a segmented image for storing the hollow wafer 10. a composite image produced;

III. A working memory 268 for storing data (eg, data and/or results) generated by a hollow wafer inspection process, such as: (a) a hollow wafer selection map indicating which of the dies 20 have been taken The segmented image related material is actually selected from the dicing wafer 5; and (b) other data, such as an error map, indicating the location or lattice position of the hollow wafer where the morphing error occurred, and possible corresponding dies Choose the type of error.

Some embodiments of SWIS 200 may also include a data storage unit (e.g., a hard disk drive). Processing unit 250 and memory 260 can be coupled to one or more other portions of the semiconductor fabrication system (eg, component sorter/die sorting device 114 and/or system control unit) via a transport/network interface unit 270 for processing The unit 250 can receive, retrieve, or access a PW map and a classification code corresponding to the hollow wafer 10 in the current consideration, and transmit the hollow wafer inspection result (eg, corresponding to the selected error map) to one or more other systems, Equipment, or device.

Typical die classification and hollow wafer inspection

In order to provide the reader with a clear understanding of the hollow wafer inspection procedure that can be managed or performed by the hollow wafer inspection module 262 of the embodiment of the present invention, various basic concepts of the hollow wafer inspection of the embodiment of the present invention will be described in detail below.

Initial setting considerations

As noted above, the remaining grains 50 typically comprise electrically poor grains, so a particular classification code can be defined or selected to correspond to poorly performing grains. In some cases, the remaining die 50 may also include die 20 that is electrically good but exhibits one or more severe visual defects (eg, lack of solder bumps or dimensional errors). Thus, one or more other classification codes can be assigned to the die that is electrically good but visually obscured and should remain on the hollow wafer 10. However, it must be understood that the definition of the remaining grains 50 depends on how the classification code is defined. For example, as needed, the remaining dies 50 are defined as those that are electrically good, such that the electrically defective dies 20 are removed and sent to a designated reject bin. If the number of electrically poor grains 20 is much lower than the number of electrically good grains, a "reversal" approach to such a grain classification operation may be more efficient and preferable. Next, the die 20 remaining on the film frame 12 (expectedly good) can be easily scraped into the bowl (another job) by a faster turntable for scanning and tape feeding operations. Prior to scraping the die 20 from the film frame 12, a hollow wafer inspection of an embodiment of the present invention may be performed to check for a grain removal error.

Prior to initiating an automated die sorting operation for one or more batches of wafers carrying the same type of semiconductor device, the operator sets or selects a die classification recipe that defines a set of classification codes. When setting up a grain classification prescription, the operator usually: (a) determines which types of dies 20 should be removed based on the combined results or possible results of the electrical and visual inspections; and (b) for electrical testing and Assign a specific classification code to the specific combined results of the visual inspection.

Dynamic update of PW map during grain classification operation

In one embodiment, the die sorting device 115 performs a set of visual inspection operations during the die sorting operation, and the visual inspection result of each die 20 is immediately updated to the PW map, and then the die 20 is simultaneously sent according to the classification code. To the intended destination. Table 2 provides a typical relationship between the PW diagram code and the classification code, including the code of the inspection job performed during the die classification job; Table 3 provides the classification code description corresponding to Table 2. According to Tables 2 and 3, the poorly-performing grains become the remaining crystal grains 50 which should remain on the hollow wafer 10. The die 20, which is electrically good and visually good, will be taken out and sent to a web (PT). Poor electrical properties and visual signs/marks, size defects, scratches, chippings, or grains of combined size and fragmentation are classified into specific boxes or plates according to the assigned classification code.

Therefore, some parts of a typical PW diagram include the data shown in Table 4 before the visual inspection during the grain classification operation:

The shaded areas of Table 5 correspond to the extracted grains that appear in a particular type of visual field in the grain classification visual inspection operation, and the PW map is updated according to the type of visual field that appears.

SWIS – Interaction of die sorting equipment

In the tandem SWIS configuration equipped with the die sorting device 115, once the die sorting operation of the hollow wafer 10 remaining on the extension stage 210 is completed, the hollow wafer inspection operation can be started to check the grain sorting operation. The removal of the effective area die 20 is correct or incorrect. The die sorting device 115 can move the docking station 210 to position the hollow wafer 10 below the SWIS image capturing device 220. As with other types of SWIS configurations, the hollow wafer 10 can be sent or shipped to a dedicated SWIS 200 in a manner that is immediately apparent to one of ordinary skill in the relevant art.

Segmentation image capture and composite image generation

One embodiment of capturing a set of segmented images after a die sorting operation is described herein. Once the hollow wafer 10 is positioned below the SWIS image capture device 220, the SWIS 200 directs the capture of at least one segmented image group of the actual hollow wafer 10 and generates at least one hollow wafer composite from the segmented image. The image will be further described below. Each segmented image corresponds to a predetermined area, portion, or segment of the entire surface area of the hollowed wafer, and/or a predetermined region, portion, or portion of a mathematical or geometric boundary (eg, a border) defined by the hollowed wafer 10 Fragment. In various embodiments, the complete segmented image set spans or encompasses 100% or substantially/substantially 100% of the hollow wafer surface area. Thus, a complete segmented image set includes at least one image, and in various embodiments includes multiple images.

After capturing the segmented image set, the SWIS 200 "splices", combines, or combines the segmented images in a digital manner to produce a composite or overall image of the hollowed wafer 10. In various embodiments, the composite image of the hollowed out wafer 10 is a virtual representation of the entire hollowed out wafer 10 (eg, 100% or substantially/substantially 100% of the hollowed wafer 10). Thus, the composite image includes image data corresponding to each grain location on the actual hollow wafer 10, including: (a) each effective region grain location, and (b) each empty shell grain location. The SWIS 200 analyzes the pixel data in the composite image to determine the grid width of the image space (ie, the number of pixels spanned by the image of the grid lines 30, 32) and the grain size of the image space (also That is, the number of pixels in the x and y dimensions spanned by the imaging of at least one grain position). As further described below, the SWIS 200 can further determine a reference origin/lattice location within the composite image; and a first/starting grain location within the composite image portion corresponding to the effective grain area of the hollowed out wafer. Next, the SWIS 200 can traverse and/or reconstruct the actual grid of the hollow wafer in the image space via a digital method, and for any particular composite image grain position, the pixel data is identified to indicate that the grain exists 50 or the grain. Vacancy 52.

Single or multiple sets of full segmentation images

In many embodiments, the SWIS 200 captures a single complete segmented image set whose effective illumination sources 230, 232 and corresponding illumination parameters in the illumination conditions facilitate the use of image processing from a single complete segmented image set. The resulting composite image reliably identifies the imaging of the grid lines 30, 32, the remaining grains 50, and the voided grain locations 52.

Alternatively, in some embodiments, multiple sets of full segmented images may be taken at the same illumination or different illumination settings for image analysis or alignment, for example, for reliable or accurate automatic determination, within a composite image Whether the pixel array or region corresponds to the imaging of the grid lines 30, 32, the remaining grains 50, and the vacant grain locations 52. Any particular complete segmented image set can be taken under the same or different lighting conditions or parameters. For example, the first component segment image can be captured under a corresponding first set of illumination conditions (eg, according to the first set of illumination intensities, the first set of illumination wavelengths, and/or the first set of illumination incident angles) Two effective illumination sources 230, 232); and the second component segment image can be taken under a corresponding second set of illumination conditions different from the first set of illumination conditions. Each set of lighting conditions facilitates the use of image processing to reliably identify the image of the ruled lines 30, 32, the remaining grains 50, and the voided grain locations 52. That is, a multi-component segment image of the hollow wafer 10 is taken, wherein each component segment image corresponds to the same, substantially the same, or overlapping region of the hollow wafer 10 under specific illumination conditions, thereby enhancing automatic identification. The confidence of the grid lines 30, 32, the remaining grains 50, and/or the voided grain locations 52 within the composite image.

Depending on the details of the embodiment, the first segment image group may be stitched together by digital means to form a first integrated image of the hollow wafer 10; the second segment image group may be stitched together by digital means to form The second integrated image of the wafer 10 is hollowed out; and the first and second integrated images may be combined or combined digitally to form a composite image. Alternatively, the first and second segmented image groups can be stitched together in a digital manner to form a composite image.

One of ordinary skill in the relevant art will appreciate that segmented image capture sequences corresponding to one or more component segment images can be determined (e.g., experimentally) for a particular type and size device fabricated in a batch of wafers, and appropriate Relevant lighting conditions or parameters. The data can be stored in a file for hollowing out wafer inspection prescription retrieval; or via a menu option input (eg, by a technician) or manual input to a graphical user interface.

In other embodiments, multiple composite images may be generated, each of which may correspond to one or more complete segmented images. Each composite image can be individually analyzed to identify the imaging of the grid lines 30, 32, the remaining grains 50, and/or the voided grain locations 52.

Cutting wafer segmentation image

In some embodiments, in addition to taking a segmented image of the hollowed wafer 10 and producing a composite image of the hollowed out wafer, the SWIS 200 can also take a segmented image of the cut wafer 5 prior to the start of the die sorting operation. And producing a composite image of the dicing wafer 5. In the above embodiment, the SWIS generates a first composite image corresponding to the dicing wafer 5 and a second composite image corresponding to the vented wafer 10. It may be helpful to establish a first composite image that corresponds exactly to the dicing wafer 5 prior to the start of the grain sorting operation, as the image may serve as a navigation aid or guide for the following operations: (a) hollow wafer inspection, advantageously Directly navigating to a particular grain location on the hollow wafer 10 after completion of the die sorting operation; and/or (b) the die classification itself, directing the pick and place device 160 during the die sorting operation. This first composite image can be used as a navigation guide for subsequent wafers in the same batch.

As explained below, the first composite image of the dicing wafer 5 not only has associated data for dicing each of the dies 20 (including the empty slabs) on the wafer 5, but also their positional information relative to the reference origin. This information may be helpful, especially as the grain sorting operation is in progress, with fewer grains 20 remaining on the hollow wafer 10 for the die sorting device 115 to reference its position in order to perform pick and place operations. In this manner, the die sorting device 115 can use the first composite image material to more easily and quickly guide the pick and place device 160 to remove individual or separate dies 20.

Depending on the details of the embodiment, the SWIS 200 may use the first composite image and/or the second composite image to spatially compare at least one composite image effective region die position to the corresponding PW pattern die position. More specifically, in determining a particular point, location, and/or die position within a composite image of a hollowed wafer, creating a composite image of the diced wafer can help improve image processing accuracy.

Once a composite image of the diced wafer is produced, the SWIS 200 can analyze the composite image of the diced wafer to determine one or more of the following: (a) the pixel width of the grid image; (b) the image of the grain 20 image (b) the position of the reference origin/lattice; (c) the position of the first/starting dies relative to the reference origin/lattice position; and (d) the corresponding area of the dicing wafer corresponding to the hollow crystal Circle the actual grid 30, 32 grid lines 30, 32 imaging layout.

The SWIS 200 may utilize one or more diced wafer composite image reference origin/lattice positions, first/start grain positions, and grid layouts to determine or assist in determining the hollow wafer composite image reference origin/ Lattice position, first/starting grain position, and grid layout. The SWIS 200 can additionally determine the hollowing out by comparing the composite image or data of the cut wafer determined from it (for example, the grid layout) with the composite image or data of the hollowed wafer determined from it (for example, the grid layout). The spread or stretch factor of the wafer relative to the dicing wafer 10. Using the data generated from the diced wafer composite image and the expansion factor, (a) improve the statistical accuracy of determining the grain position within the composite image of the hollowed wafer, for example, leaving the hollow after the grain classification operation When the die 20 on the wafer 10 is extremely small, or the film 11 of the hollowed wafer is slightly deformed due to the removal of the die 20; (b) assisting the digital image space navigation, or from the hollow wafer One grain position within the composite image traverses to another grain location; and (c) assists real/real space navigation, traversing one grain position on the hollow wafer 10 to another grain location, for example During the pick and place operation, since the grain position/grid position on the actual hollow wafer 10 has been aligned with the grain position/grid position of the composite image of the hollowed wafer, and the latter has been combined with the wafer to cut the image The grain position/grid position is aligned.

In a particular embodiment, generating a diced wafer composite image, determining a reference origin/lattice position, determining a first/start grain position, and dicing the wafer grid line 30 within the digitally reconstructed diced wafer composite image, 32, at least some of which can occur using parallel computing procedures, ie, concurrent with the die sorting operation. For the sake of brevity and conciseness, the following description of composite images of hollowed out wafers and hollowed out wafers can be similarly or equally applicable to dicing wafers and dicing wafer composite images.

For the sake of brevity and for ease of understanding, the shooting of a single complete segmented image set of the hollowed wafer 10 and the generation of a single composite image will be described below. Even so, one of ordinary skill in the relevant art will appreciate that the SWIS 200 can take images of one or more components of the hollow wafer 10, as well as image one or more segments of the wafer 5 to produce at least A composite image in a manner substantially equivalent or similar to that described below.

Identification of effective area grain position of composite image

As described above, the PW map presents only a small portion of the total surface area of the wafer, and a small group of composite images of the wafer 10 is hollowed out. However, the composite image of the hollowed wafer 10 is typically an image of the entire wafer that is actually hollowed out, spanning the effective area die position and the empty/mirror grain position. Therefore, the PW pattern grain position corresponding to the wafer effective area grains may be mapped to a plurality of areas within the composite image. Therefore, between the die position of the PW pattern and the die position group corresponding to the effective area die position in the composite image of the hollowed wafer, it is necessary to solve the problem of spatial alignment or registration.

To address this issue, the SWIS 200 spatially aligns, registers, aligns, or identifies the hollow wafer composite image portion corresponding to the effective region die position and the PW map effective region die position.

Determination of the reference point of the hollow wafer reference or reference grid

It should be noted that information known to the SWIS 200 or SWIS 200 can be provided, including wafer size, grain size, and dicing wafer orientation on the film frame 12. The orientation of the composite image relative to the film frame 12 is also known. In various embodiments, in order for the SWIS 200 to spatially correctly align each active area die position within the composite image of the hollowed wafer with its corresponding effective area die position on the actual hollow wafer 10 and in the PW map The SWIS 200 determines by analyzing a composite image of the hollowed out wafer: (a) a reference origin, and/or (b) a reference grid position. The reference origin or reference grid position corresponds to a pixel coordinate within the composite image, which itself corresponds to a known or predetermined source, point, or grid location on the actual hollow wafer 10. A reference origin within a composite image can be a specific point in the image space or pixel space. The reference grid position can be a specific point in the image space or a pixel array at a specific location in the image space. Thus, the reference grid location itself may be similar or substantially equivalent to or defined as a reference origin.

According to the reference origin/lattice position, the grain position of each effective region in the composite image of the hollowed wafer can be determined, including the first/starting crystal corresponding to the predetermined first/starting grain position in the PW pattern. Grain position. The first/starting grain position within the composite image corresponds to a particular grain position within the effective grain area of the actual hollowed wafer, its distance from the reference origin and its relative direction are known. Within the composite image, the first/starting dies are located in the image space at a particular image spatial distance and relative direction from the reference origin/lattice position. Once the first/starting grain position in the composite image is determined, the composite image data corresponding to the grain position of the effective area of each composite image can be compared with the corresponding PW image data to check the grain 20 The effective area of the actual hollowed out wafer has been removed correctly or incorrectly, as further described below.

The reference origin/grid position is similar to the reference die 21 of the die sorting operation, but in various embodiments of the invention, the reference origin/lattice position corresponds to the hollow wafer or die position, rather than a reference The position of the die 21. That is, the reference origin/lattice position corresponds to the hollow wafer position or grain position, which is different from the position of the reference die 21.

The typical way to determine the reference origin

The reference origin in the composite image of the hollowed wafer and the relative position of the first/starting die can be determined in a variety of ways. In embodiments where the hollow wafer 10 generally assumes a regular shape, the reference origin may be defined as the center point of the hollow wafer 10. For example, most wafers are circular, so the SWIS 200 can determine the reference origin by identifying at least three points within the composite image using conventional image processing techniques, which are located at the hollow wafer 10 Around. In accordance with these points, conventional image processing techniques can determine the perimeter of a circle having the same diameter as the hollowed wafer 10, thereby determining the center point of the hollowed out wafer (e.g., a pixel coordinate in the image space). Even if the surrounding portion of the processed wafer is missing due to processing or processing operations, the determination of the center point of the hollowed out wafer will not be affected because the curvature of the hollowed out wafer can be inferred using algorithms and is highly accurate in a manner relevant Those of ordinary skill in the art will immediately understand. Once the center point of the wafer 10 is found in the composite image, the relative position of the first/starting grain in the composite image can be immediately identified because the position of the first/starting die relative to the reference origin It has been determined in advance according to the number of pixels or the distance and the horizontal and vertical grid distance (for example, considering the resolution of the SWIS image capturing device).

In some embodiments, the position of the first/starting die relative to the reference origin is determined by identifying a set of grid lines 30, 32 within the composite image based on the known actual layout of the processed wafer. And moving to the first/starting grain position within the composite image by calculating the number of lateral and longitudinal grid lines 30, 32 of the known first/starting grain position distance reference reference origin. In the above embodiment, SWIS 200 may utilize image processing algorithms to identify grid lines 30, 32 within the composite image. In the image space, the grid lines 30, 32 are horizontal and vertical narrow lines spanning a predetermined number of pixels (eg, 3 - 5 pixels, depending on the resolution of the SWIS image capture device), corresponding to the actual grid lines width. Since the grain size is known, the lateral and vertical pixel sizes of each grain position within the composite image are also known. Depending on the grain size, the predetermined grid lines 30, 32 within the composite image can be determined such that the relative position of the first/starting grain location and each of the other effective grain locations within the composite image can be determined immediately.

Although in some embodiments, the reference origin is defined as hollowing out the center of the wafer 10 (or cutting the wafer 5 or the center of the processed wafer), in other embodiments this may not be the case, they may select a suitable crystal. A circle feature defines or identifies a reference origin. The wafer characteristics used to define the reference origin depend on the actual characteristics of a batch of wafers in question and whether the feature can be easily and consistently identified within the composite image. The reference origin can also be selected based on the information provided by the semiconductor wafer manufacturer, that is, the specification of the actual actual wafer specifications for a particular size and size of device (eg, integrated circuit). This material is usually included in a document commonly referred to as the "actual wafer standard document" and is available from various wafer manufacturers in a manner that is immediately known to those of ordinary skill in the relevant art.

Figures 8A and 8B show portions of a typical actual wafer standard file. The actual wafer standard document provides information on the widest lateral and vertical dimensions of the wafer, as well as the location of the reference die 21 for fabricating a particular type and size of device on a particular size wafer. From a specific actual wafer standard document, an engineer or technician can define or define a horizontal reference line and a longitudinal reference line, and the intersection point can be defined as a reference origin to determine the position of the first/starting die. (and/or the position of the wafer reference die 21). The SWIS 200 can analyze the composite image to identify pre-determined or predefined lateral and longitudinal reference lines. For example, data indicating which grid lines 30, 32 are defined as horizontal and vertical reference lines can be included in the prescription for the hollow wafer inspection.

Additional or additional information may also be used to define the reference origin. For example, in another embodiment, the bottommost lateral grid line 30 on the wafer closest to the wafer ID can be used as a lateral reference line, and a particular longitudinal grid line 32 (eg, intersecting the lateral reference line) The leftmost or rightmost vertical ruled line 30) can be defined as a longitudinal reference line. The intersection between the lateral and longitudinal reference lines may be defined as a reference origin relative to the predetermined position of the first/starting dies.

Typical way to determine the position of a reference grid

In addition to determining the reference origin, some embodiments may identify or reconstruct the grid layout of the hollowed wafer from the composite image via a digital manner and determine a reference grid point or location within the composite image. The position of the first/starting die can then be determined relative to the reference grid position.

In order to digitally identify/reconstruct the grid layout of the hollowed out wafer, the SWIS 200 uses conventional image processing techniques to analyze the composite image. As described above, the imaging of the grid lines 30, 32 can be identified by the intensity of the pixels within the narrow or very narrow rows of pixels, while the pixel intensity in the larger or significantly larger pixel regions corresponds to the remaining The grain 50 and the voided grain location 52. The SWIS 200 can identify/reconstruct each horizontal grid line 30 and longitudinal grid line 32 within the composite image via a digital manner.

As will be appreciated by those of ordinary skill in the relevant art, one or more rows of grains 20 may be detached from the wafer due to processing or processing. Thus, the total number of lateral and/or longitudinal grid lines 30, 32 within the composite image may be less than the total number of lateral and/or longitudinal grid lines 30, 32 on the actual wafer. In some embodiments, SWIS 200 can calculate the total number of horizontal and vertical grid lines 30, 32 within the composite image and analyze the outermost top, bottom, left, and right k-line lines 30, 32 (eg, k = The absolute or relative length of 3, 4, or 5), and the absolute or relative length is compared with the known grid layout of the wafer, and one or more rows of outer rows missing from the hollow wafer 10 are identified based on the composite image. Grain 20.

If the grid lines 30, 32 reconstructed in a digital manner conform to the known grid layout of the wafer, the SWIS 200 may define a certain center point of the grid reconstructed in a digital manner as a reference origin, which is located in the reference grid position or It is adjacent (for example, the most central or central grid position). The SWIS 200 may calculate a predetermined number of lateral and longitudinal grid lines 30, 32 of the first/starting dies from the reference grid position in a predetermined direction, thereby determining the position of the first/starting dies relative to the reference grid position. If the SWIS 200 determines that one or more rows of the dies 20 are missing from the hollow wafer 10, the SWIS 200 can make appropriate lateral and/or longitudinal directions when determining the position of the first/starting dies relative to the reference grid position. The grid is offset to accommodate the missing grains 20 rows.

In some embodiments, the SWIS 200 may additionally or alternatively identify or digitally reconstruct the grid lines 30, 32 via identifying the longitudinal and lateral edges of the die 50 left on the hollow wafer 10. Since the ruled lines 30, 32 are adjacent and parallel or perpendicular to the edge of the die, and extend across the entire hollow wafer 10, and the known predetermined line width is known or known, the grid reconstruction is simple and easy, but at least A small number of dies 20 remain in the effective area of the hollow wafer corresponding to the grain positions of the plurality of different grid lines 30, 32.

Move to the first/starting die position and/or other die positions in the image space

In various embodiments, to verify that there are 50 or grain vacancies 52 (e.g., effective area die positions) for the composite image grain locations corresponding to the actual grid locations on the hollow wafer 10, the SWIS 200 is continuously analyzed or A pixel array corresponding to the position of the composite image grain is evaluated. The SWIS 200 continuously moves or faces, traverses, and/or passes through the composite image die position in accordance with the image space or pixel space size of each die 20. The image space grain size is determined in advance, and is compared with the actual grain size and the resolution of the SWIS image capturing device 220, and can be confirmed by analyzing the composite image by SWIS.

Once the SWIS 200 identifies the reference origin and/or reference grid position of the composite image, in many embodiments, the SWIS 200 can be moved or transitioned to the first/starting die position within the composite image by: : (a) determine or confirm the shortest distance from the reference origin/lattice position to the first/starting grain position (eg along the shortest diagonal); (b) determine or confirm the reference origin/grid position a predetermined or shortest x-y grain position to the first/starting grain position via the grain position path; and (c) traveling along the grain position along the grain position path by individual grain positions to reach the first One / starting grain position. When the SWIS 200 travels along the path to each die position, the SWIS 200 can analyze the pixel array data corresponding to each die position to verify whether each predetermined die position is indeed correct and in the composite image. The grain positions are aligned.

In other embodiments, the SWIS 200 may calculate or confirm the number of x-direction pixels and the number of pixels in the y-direction between the reference origin/lattice position and the first/starting grain position, and by moving within the composite image The appropriate number of pixels in the x and y directions relative to the reference origin/reference grid position is moved directly from the reference origin/lattice position to the first/starting grain position.

Avoid systemic errors

A significant difference between the various embodiments of the present invention and the prior art for automated hollow wafer inspection is that when determining the reference origin or reference grid position and the first/start grain position, the SWIS 200 The same reference die 21 that is automatically identified prior to the start of the die sorting operation or manually identified/detected by the operator is not used. The reason for this is to avoid repeating any systematic errors that may occur as a result of the operator identifying the faulty die as the reference die 21. If the operator has identified the wrong die as the reference die 21, a systematic die take-out error will occur, and the hollow wafer inspection operation performed using the same incorrect reference die 21 will therefore not be detected or recognized. Any grain removal error. In order to avoid the above systematic errors, the SWIS 200 refers to another set of features on the actual wafer to define a reference origin or reference grid position with respect to which the other crystal grains in the effective area of the composite image of the hollow wafer can be determined. Location, including the first/starting die.

The definition of the reference origin and/or reference grid position can be pre-determined for the specific type and size of the device manufactured by the entire batch of wafers, and the data can be stored in the file for starting the die sorting and hollowing out the wafer inspection. Before the operation, it is used for grain classification and hollow wafer inspection prescription retrieval. Alternatively, the material can be entered via a menu option (eg, operated by a technician) or manually entered into a graphical user interface.

Comparison of data content between composite image and PW image

After the above spatial comparison, alignment, or registration, the data content comparison can be performed, and the SWIS 200 compares the following data contents during the period: (a) image data corresponding to the effective area grain position in the composite image. And (b) the data of the corresponding grain location generated by the PW diagram or generated using the PW diagram (eg, the updated data by any final visual inspection performed by the grain classification operation). The above data content comparison helps to automatically determine that the die 20 is correctly or incorrectly removed from the active area of the actual hollow wafer during the die sorting operation.

Within the composite image of the hollowed out wafer, a pixel array corresponding to the grain position of each active region can provide captured image data (eg, pixel values) indicating the presence of grains on the actual hollowed wafer 10 50 Or the wafer vacancy 52 of the die position is actually hollowed out. Each of the pixel arrays described above has a pixel area that corresponds to the actual size of the die 20 and depends on the resolution of the SWIS image capture device 220 in a manner that is immediately apparent to those of ordinary skill in the relevant art.

In various embodiments, conventional image processing techniques are utilized to analyze each pixel array corresponding to the effective area grain position within the composite image of the hollowed wafer to determine if the grain position has indicated that the grain is present 50 or crystalline. Grain vacancies 52. For example, the grain presence 50 can be indicated by a pixel array corresponding to a particular effective grain location and having a first average intensity; and the die vacancy 52 can be indicated by a pixel array having a second average intensity And the second average intensity is significantly different from the first average intensity. For the specific composite image effective region grain position in consideration, the pixel array analysis result is compared with the PW map data of the grain position, and includes a code indicating that the corresponding actual die 20 should have It is taken out from the dicing wafer 5 or should be left on the stencil wafer 10. If there is an inconsistency between the composite image pixel array analysis and the PW map data for any particular effective area grain position, a fetch error has occurred.

More specifically, in some embodiments, for each effective region die position within the composite image, SWIS 200 directly compares the mathematical values corresponding to the values of the composite image pixel array of the hollowed wafer and the die position The PW map code and the updated PW map using any visual inspection performed during the die sorting operation. This may involve SWIS generating a hollow wafer selection map based on the composite image pixel array values, where the digit code corresponding to the grain position of each active region indicates that the grain exists in the composite image of the hollowed wafer, or from the hollowed out The wafer composite image is vacant. In some embodiments, the SWIS 200: (a) after the designation/selection of the classification code, the hollow wafer selection map corresponding to the dicing wafer 5 is generated from the PW pattern before the dies 20 are actually removed from the dicing wafer 5; And (b) comparing the pixel array mathematical value corresponding to the grain position in the composite image of the hollowed wafer or the hollow wafer selection map with the content of the hollow wafer selection map corresponding to the die position.

Based on the data content comparison between the partial composite image and the PW image, the SWIS 200 can determine: (a) the die 20 that should have been taken out of the dicing wafer 5, and thus should have been vacant from the vacant wafer 5, whether or not Properly remaining on the hollow wafer 10; and/or (b) should not be removed from the dicing wafer 5, so the dies 20 that should appear on the vented wafer 5 are vacant from the vented wafer 10.

Automated hollow wafer inspection program

In view of the above, some flow charts will be described below to detail a hollow wafer inspection procedure in accordance with an embodiment of the present invention. 9 is a flow diagram of an automated process 300 for hollow wafer inspection; FIGS. 10A-13 are flow diagrams of automated hollow wafer inspection oriented processes in accordance with certain embodiments of the present invention. The specific hollow wafer inspection operation described in FIGS. 9-13 can be performed by a tandem SWIS 200 that cooperates with a die sorting device 114 to receive a hollow wafer via a common expansion stage 210, the expansion stage being a die The sorting device 114 is shared with the SWIS 200; or by a hub or stand-alone SWIS 200, which has its own extension station 210, which is separate from the extension station of the die sorting device 114.

Typical setting operation

In FIG. 9, the general automation program 300 includes a first program portion 302 that involves performing a hollow wafer inspection setting job involving setting/arranging a recipe selection, retrieval, or definition, such as a composite image generation prescription, a grain classification prescription. And/or hollow wafer inspection prescriptions. Therefore, for each batch of wafers with the same type of semiconductor device, before the start of the automatic die sorting operation, the operator usually sets the prescription for the die sorting operation: (a) based on the combination of electrical testing and visual inspection. The result or possible outcome determines which type(s) of the die 20 to remove; and (b) assigns the appropriate classification code for the specific combined results of the electrical test and the visual inspection. The appropriate classification code is assigned to determine that each active area die 20 should: (a) be selected from the dicing wafer 5 and sent to a particular sort destination, resulting in a vacancy in the die 10 on the wafer 10; or (b) It remains on the hollow wafer 10 and becomes the remaining crystal grains 50.

In one embodiment, the SWIS 200 is coupled in series with a die sorting and visual inspection system 114, which can be presented to the graphical user interface as a visual object (e.g., an icon) that can be selected by a technician or operator. In response to the SWIS setup/configuration program selection, a software module can be executed by which a technician or operator can select or define a prescription that can be defined as a composite image generation prescription or a hollow wafer inspection prescription. Depending on the details of the embodiment, the composite image generation/snap wafer inspection prescription may be an independent prescription or incorporated into a grain classification recipe. Thus, the die classification and hollow wafer inspection recipes can include data corresponding to conventional grain classification recipes, as well as additional data used to generate composite images that help to hollow out wafer inspections. For tandem SWIS 200, which is used to inspect the entire batch of hollow wafers after the die sorting operation, the hollow wafer inspection prescription should normally be established at the same time as the die sorting job prescription.

Setting/configuring a prescription can include at least a few of the following materials:

(a) wafer lot data and each wafer ID in the batch in the consideration;

(b) actual wafer data, such as: (i) wafer size; (ii) grain size; (iii) wafer grid data, including the number of horizontal grid lines 30 corresponding to the grain columns and corresponding to the grain rows The number of vertical grid lines 32;

(c) SWIS image capture device selection:

(i) depending on the resolution of the image, the field of view of the camera relative to the grain size;

(ii) the degree of segmentation image overlap;

(iii) the number of segmented images to be captured in the case of known wafer size and field of view and resolution of the image capture device;

(d) illumination source selection and corresponding illumination parameters;

(e) a reference origin and/or a reference grid position definition relative to the wafer grid position (eg, a lateral reference line, a position of the longitudinal reference line);

(f) a first/starting grain coordinate on the actual wafer grid relative to the reference origin and/or the reference grid position;

(g) actual reference to the predetermined auxiliary reference die position and/or reference feature position on the dicing wafer, relative to the reference origin position on the dicing wafer, reference grid position, and/or first/starting crystal Grain, in the form of relative encoder position offset or relative (x, y) coordinate offset;

(h) the left, right, top, and bottom or peripheral edges of each dicing wafer, the outermost or outermost k-line groups (eg, 3, 4, or 5 grid lines) relative to the glue during the cutting The absolute and/or relative length of the orientation on the membrane frame 12.

Local PW map soundness inspection

Referring again to FIG. 9, in one embodiment, the automated hollow wafer inspection process 300 includes a second program portion 304 that relates to verifying the soundness of the local PW map used by the die sorting device 114 in visual inspection and classification operations.

As explained above, the PW map is a data archive or data set containing the following data: (a) the grid position of the die 20 for each active area on the actual wafer; (b) the electricity for each of the die 20 described above. The result of the sex test and the visual inspection code; and (c) a classification code (destination/box number) assigned according to the electrical test and visual inspection result of each die 20. This information is typically stored in a predetermined data file format or database format, such as in a database of host servers remote from the die sorting device 114 and SWIS 200.

In general, a copy of a PW map for each particular wafer in the wafer lot in question is retrieved from the host server before starting the die sorting operation. The method is to separate the host PW map data from the host server, and locally reconstruct and store the data in the die sorting device to become a partial PW map. The manner in which data shunted from the host server is decoded and ordered in the local PW map depends on the decoding protocol. There is usually a contract for distributing and sorting data from the host server repository. In the semiconductor industry, the protocol used for shunting (data strings) of the data on the wafers 20 on the wafer is called the SECS/GEM (SEMI Equipment Communication Standard/General Equipment Model) agreement. However, there are eight variants of this agreement, and each variant version affects how the ordered datasets are arranged within the local PW diagram. In addition, not every semiconductor manufacturing equipment complies with this agreement. When deviating from the particular protocol or protocol variant version used to offload or sequence and decode the host PW map material, the received data will be presented in a different manner than expected; there will also be a translation error. Errors in decoding host PW map data using incorrect protocols or protocol variants will mean that some of the dies 20 and all of the relevant data for those dies 20 may have been incorrectly presented in the local PW map against the actual wafer. . An error will occur when the grain 20 grid position presented by the local PW map is compared to the actual wafer (and host PW map).

This will have two consequences. First, the pick and place operation performed in accordance with the erroneous partial PW map may result in the erroneous die 20 being removed from the dicing wafer 5 and/or updating the inspection result to the erroneous grain position of the local PW map. Second, it will not be possible to detect the fetch error by simply aligning the actual wafer 10 with the local PW map because the die sorting device 114 will correctly perform the error indication.

In various embodiments, the SWIS 200 will check the sanity of the local PW map prior to performing further hollowing inspection operations. Alternatively, the die sorting device 114 may check the soundness of the local PW map before performing the die sorting operation. That is, depending on the details of the embodiment, the check procedure for the soundness of the partial PW map can be used as a part of the SWIS inspection program or as a part of the grain classification and inspection operation. For the sake of brevity and help in understanding, the soundness of the local PW map checked by the SWIS 200 will be described below. However, one of ordinary skill in the relevant art will immediately appreciate that the sanity check may also be performed by the die sorting device 114 prior to performing the die sorting operation to avoid fetch errors due to lack of soundness of the local PW map.

FIG. 10A is a flow diagram of a typical first routine 400 performed by SWIS 200 to verify the health of a local PW map in accordance with an embodiment of the present invention. In the first program portion 402, the host server generates a first data set according to its database content in accordance with a standard file format to present a host PW map. For the sake of brevity, the first data set is defined herein as the first ASCII file. Therefore, the first ASCII file represents the host PW map. In the second program portion 404, the host server stores the first ASCII file in a location accessible to the SWIS 200. The first ASCII file is based on the electrical test results provided by the host-side wafer result database, indicating the position or grid position of each of the electrically defective grains on the processed wafer or the diced wafer 5. In the first ASCII file, the defective dies at each location on the processed wafer can be indicated by a digital or digital code/value, such as "1", while other die positions can be encoded by another digital or digital code. The value indicates, for example, "0". The first program 400 additionally includes a second program portion 404 that relates to the host system storing or transmitting the first ASCII data set to a location, address, or device (e.g., network drive location) accessible to the die sorting device 114. .

FIG. 10B is a flow diagram of a typical second routine 410 performed to verify the health of a local PW map in accordance with an embodiment of the present invention. The second sanity check procedure 410 can be performed by the SWIS 200 without translating between the host side data format and the die sorting device data format. In various embodiments, the second sanity inspection program 410 includes a first program portion 412 that involves the SWIS 200 retrieving or entering a partial PW map of the processed wafer in consideration; and a second program step 414 involving generating a second ASCII file. In the second ASCII file, the location of the defective die on the processed wafer can be indicated by a digital or digital code or value, such as "1", while other die positions can be represented by another digital or digital code or value. Indicates, for example, "0". The digital or digital code/value type for the location of the electrically defective die in the second ASCII file corresponds to or conforms to the digital or digital code/value type for the location of the defective die in the first ASCII file.

The second sanity check program 410 further includes a third program portion 416 that involves the SWIS 200 retrieving the first ASCII file from the location or address at which the host system stores the first ASCII file; the fourth program portion 418, which relates to the SWIS 200 comparison first. The data content of the ASCII file and the second ASCII file; and the fifth program portion 420, which relates to the SWIS 200 determining whether the first and second ASCII files match in the same position indicating that the wafer is not properly printed on the wafer. The fourth program portion 418 can involve one or more mathematical tasks, such as subtracting another data set from one data set, as will be immediately apparent to one of ordinary skill in the relevant art.

If the first and second ASCII files match the corresponding die positions of all the defective dies on the processed wafer in the consideration, the sixth program portion 422 confirms that the local PW map is sounder relative to the host PW map. And/or accepting a partial PW map such that further work can be performed or continued in accordance with the local PW map, including hollowing out wafer inspection operations.

If the first ASCII file and the second ASCII file do not match the same grain position indicating that the wafer is poorly grown, the local PW pattern in the consideration is damaged. The second sanity check program 410 can accordingly include a seventh program portion 430 that involves the SWIS 200 displaying the partial PW icon as incorrect, and an eighth program portion 432 that involves the SWIS 200 transmitting a process map error message or notification to the host system. Depending on the details of the embodiment, the pattern error message may include a second ASCII file or reference to a host access location (eg, a network drive location) with reference to the second ASCII file for further error analysis.

The second soundness verification program 410 can also include a ninth program portion 434, which relates to the SWIS 200 transmission classification code group to the host system; and a tenth program portion 436, which relates to the host system wafer result data of the processing system based on the host system. The library data and the received classification code produce a corrected partial PW map. In various embodiments, the corrected local PW map is an ASCII archive indicating that the grain position of each processed or diced wafer indicates that the die 20 remaining at the grain location should: (a) avoid removal, resulting in hollowing out The grain position of the wafer 10 remains on the wafer 10; or (b) is removed, resulting in a grain location 52 that is vacant on the wafer 10. In the corrected local PW diagram, corresponding to the electrically good crystal grain 20, the grain position or lattice position to be taken out may be represented by a digit code or numerical value, for example, “0”; and corresponding to the electrically defective crystal grain 20, The grain position or grid position that should be avoided should be indicated by another digit code or value, such as "1". Finally, the second soundness verification program 410 further includes an eleventh program portion 438 that involves the host system transmitting the corrected partial PW map to the die sorting device 114; or the host system stores the corrected partial PW map to a certain Enter the location, address, or device of the SWIS 200, and transmit a corresponding message or notification to the SWIS 200.

Segmentation image capture and composite image generation

The automated hollow wafer inspection process 300 of FIG. 9 further includes a third program portion 310 that involves positioning the hollow wafer 10 below the SWIS image capture device 220. In an embodiment in which the SWIS 200 is integrated or in series with the die sorting device 114, the third program portion 310 involves moving the die sharing device 114 and the SWIS 200 common extension station immediately after the die sorting device 114 completes the die sorting operation. 210 to a predetermined position below the SWIS image capturing device 200. In an embodiment where the SWIS 200 is not in series with the die sorting device 114, the third program portion 310 involves providing at least one hollow wafer 10 to the SWIS 200, such as via a box containing the film frame 12 in which the hollow wafer 10 is located; The film frame 12 to the SWIS extension 210, for example via a robotic arm; applying a vacuum to the film frame 12; re-stretching the film 11 to a predetermined amount; and moving the position of the extension table 210 such that it is placed on the SWIS A predetermined position below the image capturing device 220.

As shown in the fourth program portion 312 of FIG. 9, once the extension station 210 is positioned below the SWIS image capture device 200, the SWIS image capture device 200 can initiate the first and/or in accordance with the particular illumination parameters described above. The second set of illumination sources 230, 232, then when the extension station 210 is displaced in accordance with the predetermined segmentation image capture mode as shown in the fifth program portion 314 of FIG. 9, the SWIS 200 can begin to take a segmentation map of the hollow wafer 10. image.

11A and 11B are schematic illustrations of a typical method of capturing a set of segmented images 900a-i of a hollow wafer 10 by SWIS image capture device 220 to produce a composite image 1000, in accordance with an embodiment of the present invention. In various embodiments, each segmented image 900a-i corresponds to a particular region or portion of a mathematical or geometric perimeter, such as a bezel 950 that encloses (eg, wraps around and is slightly larger than) the hollow wafer 10. For example, in an embodiment for capturing a plurality of segmented images 900a-i, each segmented image 900a-i may correspond to a predetermined segment of the border area, such as 1/36 of the border area, 1/ 25, 1/16, 1/9, 1/4 or 1/2.

Each segmented image 900a-i includes image material corresponding to a particular location of the hollow wafer 10 relative to the SWIS image capture device 220 in a manner that is immediately apparent to those of ordinary skill in the relevant art. Depending on the wafer size, grain size, and the position of the circular/substantially circular hollow wafer 10 relative to the position of the SWIS image capture device 220, the image data included in each of the segmented images 900a-i may be Corresponding to many grains 20, such as tens, hundreds, or thousands of grains 20. The field of view of the image capture device 220 need only be sufficient to capture each segmented image 900a-i. As shown in FIG. 11A, the first illumination source set 230 is a portion of the space used to direct illumination across the underside of the cut wafer, which is typically slightly larger (at least as large) as a predetermined segment of the border 950 of any segmented image ( Corresponds to the field of view of the SWIS image capture device).

In general, the number of segmented images 900a-i to be captured depends on a number of coefficients, such as: (a) wafer size; (b) grain size; (c) characteristics of the SWIS image capture device 200, such as analysis Degree (usually fixed) and field of view (usually adjustable); and (d) image processing to accurately and reliably determine the composite image resolution required for grain presence 50 or grain vacancy 52 at each lattice location. In the exemplary embodiment shown, a total of nine segmented images 900a-i are displayed, such that each segmented image 900a-i corresponds to 1/9 of the total area of the bezel 950. In some embodiments, the segmented image set 900 is a single image, that is, a segmented image itself directly acts as a composite image of the hollowed out wafer 10.

Once the SWIS 200 has taken the segmented image set 900, the SWIS 200 can stitch the segmented images 900 through a digital manner or algorithm to form a composite image 1000, as shown in the sixth program portion 320 of FIG. . To facilitate segmented image stitching, in various embodiments, each segmented image 900 overlaps with its neighboring closest segmented image 900 (eg, above, below, and adjacent to the segmented image 900 in question). a predetermined amount. The degree of overlap depends on the accuracy with which the image processing algorithm can identify any two overlapping overlapping grain features (e.g., grain edges) that are closest to the segmented image 900 such that the common grains of these overlapping segmented images 900 Features can be reliably matched. Usually the degree of overlap is equal to or approximately equal to one grain size, but it is not necessary to overlap this size. If one or more portions of the die 20 have sufficient features such that the grain edges, boundaries, structures, or features within a segmented image 900 can overlap the same grain edge that is closest to the segmented image 900 If the boundaries, structures, or features are properly aligned, the degree of overlap can be smaller. The degree of overlap depends on the extent to which the overlapping pixels corresponding to the same features on the same die 20 can exactly match, while in some embodiments, the segmented image overlap can be less than one grain size.

In addition to the above, when each segmented image 900 is captured, the segmented image 900 can be stitched in a digital manner using encoder data corresponding to the relative position of the extension table 210 within the perimeter of the bezel 950. Based on the encoder data, the SWIS 200 can: (a) determine the center of each segmented image 900 and the relative distance traveled between each segmented image 900, and (b) digitally place each The segmented images 900 are positioned and stitched or combined. Strictly speaking, this technique is not required for segmentation image overlap. However, in order to achieve greater precision, in some embodiments, the SWIS 200 can combine (a) an imaging processing algorithm and (b) an extended station encoder data in a digital manner when capturing each segmented image. The segmented images 900 are stitched together. In the above embodiment, the degree of overlap between the segment images 900 can be reduced.

Typical composite image generation program

Figure 12 is a flow diagram of a typical composite image generation program 500 in accordance with an embodiment of the present invention. In one embodiment, composite image generation program 500 includes a first program portion 502 that involves selecting a first segmented image 900a, and a second program portion 504 that involves selecting one or more adjacent first segmented images. Additional segmented image 900b, f. The third program portion 506 relates to identifying surrounding reference features, such as grid lines and/or grain edges or boundaries shared by adjacent segment images 900a, b, f via image processing jobs (eg, along a certain segment) Multiple grain edges of the grain matrix). The fourth program portion 508 involves registering or aligning adjacent segment images 900a, b, f in consideration along a shared or common surrounding reference feature in a digital manner. The fifth program portion 510 relates to storing the registered segmented image groups 900a, b, f as intermediate images, and a sixth program portion 512 relating to determining whether additional segmentation image analysis is required relative to the intermediate image. And registration/alignment. If so, the composite image generation program 500 returns to the second program portion 504 to select one or more other segmented images 900 that are closest or adjacent to the intermediate image such that the one or more additional segmented images 900 can be Align and blend into the middle image. Once a set of segmented images 900 spanning the entire wafer dicing wafer or hollowed out wafer surface area is considered, the composite image generation program 500 specifies or stores the most current intermediate image as a composite image.

After splicing the segmented image 900, an accurate composite image 1000 of the entire stenciled wafer 10 is formed, which accurately presents each of the remaining dies 50 on the vented wafer 10 and the die position of each vacancy The relative position of 52.

Find the first/starting grain position in the composite image

In various embodiments, before the pixel region of the composite image 1000 can be analyzed for the corresponding grain position relative to the local PW map data, the SWIS 200 first determines that the composite image pixel region represents the first/start The predetermined pixel location or area of the die. Thus, referring again to Figure 9, the routine 300 includes a seventh program portion 330 that relates to identifying a first/starting die position within the composite image.

The predetermined first/starting grain pixel position may correspond to the center of the first/starting die or a particular corner of the first/starting die. The predetermined pixel location may be determined via one or more of the above methods, such as by the use of image processing techniques, in a digital manner: (a) identifying a reference origin within the composite image 1000 (eg, corresponding to the hollow wafer 10) a center, or the intersection of the lateral reference line and the longitudinal reference line), and determining a predetermined position of the first/starting die relative to the reference origin; and/or (b) reconstructing the lattice layout of the hollowed wafer and reconstructing therefrom The layout identifies the reference grid location and determines a predetermined location of the first/starting die relative to the reference grid location.

Analysis of composite images and identification of crystal removal errors

Once the pixel position of the first/starting grain in the composite image 1000 is determined, the SWIS 200 can analyze the pixel region corresponding to the grain position of each effective region in the composite image 1000, as shown in FIG. Program portion 340 is shown. Based on this analysis, the SWIS 200 can be determined for each of the above-described die positions, and the composite image is indicative of a grain presence 50 or a grain void 52 on the hollow wafer 10. In various embodiments, for each pixel region analysis of the effective region grain position, the SWIS 200 will generate a hollow wafer pick pattern with an indication of the presence or grain vacancy indication or code. In this manner, the hollow wafer pick pattern is indicated or coded for each active area die position, either that the remaining die 50 is present on the hollow wafer 10 or that the die location is vacant 52. The hollow wafer selection map can be a data file, such as an ASCII file, which is arranged in a predetermined format so that the contents of the data can be effectively compared with the data in the local PW map.

After the SWIS 200 generates a hollow wafer selection map and indicates the presence of the die and the vacancy indication for each effective region die position, the SWIS 200 can compare the wafer selection map and the local PW map data for the above-described die position comparison. To determine if any grain removal errors have occurred at these die locations, as shown in the ninth and tenth program portions 350, 352 of FIG.

13 is a flow diagram of a typical process 600 for generating a hollow wafer selection map and identifying a die removal error in accordance with an embodiment of the present invention. In one embodiment, the routine 600 includes a first program portion 602 that relates to selecting a first/starting die location or a next/continuous die location within the composite image 1000 (eg, directly adjacent die locations) . Selecting the first/starting grain position may include selecting a known pixel location within the composite image pixel region representing the first/starting grain, such as a pixel location corresponding to the first/starting grain center, Or the predetermined corner of the starting die. Selecting the next/continuous die position may include shifting a predetermined image space die separation distance from the current or most recent composite image die position, that is, a predetermined number of pixels, the moving direction corresponding to the effective region die position : (a) the actual arrangement on the wafer, and (b) the sequencing or sequencing in the local PW diagram.

In general, the predetermined pixel distance defining the inter-die separation in the image space corresponds to the actual grain size and the actual width of the trench-cut trench or grid lines 30, 32, and is based on the actual wafer. The layout, the extent to which the film 11 carrying the dicing wafer 5 is stretched, and the resolution of the SWIS image capturing device 220. As will be appreciated by those of ordinary skill in the relevant art, the actual width of the hollowed wafer trench or grid lines 30, 32 is greater than the actual width of the uncut wafer grids 30, 32 due to the stretch of the film 11. In various embodiments, the pixel width of the hollowed-wafer grid lines 30, 32 can be determined via conventional image processing techniques, for example, in conjunction with the determination of the pixel position of the first/starting grain within the composite image 1000. stand up.

The program 600 includes a second program portion 604 that relates to analyzing pixel intensities corresponding to the currently selected grain position within the pixel region and mathematically determining the average, median, or other statistically significant of the pixels within the pixel region. Strength of. The third program portion 606 is operative to determine that the current grain position is a grain presence 50 or a grain vacancy 52 based on the mathematically determined intensity. As described above, the grain presence 50 will correspond to the first average intensity, while the grain void 52 will correspond to the second average intensity, and the second average intensity will be significantly different (eg, greater than) the first average intensity.

The fourth program portion 608 relates to storing a grain presence indication or a grain vacancy indication at a hollow wafer pick pattern location corresponding to the grain position in the consideration. For further understanding, see Figures 7A and 7B and Figure 7C again. As described above, FIG. 7A shows a typical composite image 1000 of the hollowed out wafer 10; FIG. 7B shows an enlarged portion 1002 of the hollowed-wafer composite image 1000, which corresponds to a 2 x 7 array of hollowed-wafer lattice locations, It is defined by horizontal and vertical grid lines 30, 32. FIG. 7C illustrates a typical digitally encoded code within a typical partially hollowed wafer selection map 1100 that corresponds to the enlarged portion 1002 of the composite image 1000.

More specifically, in the enlarged portion 1002 of the composite image 1000, a total of five composite image lattice positions indicate that the crystal grains exist 50; a total of nine composite lattice positions indicate the grain voids 52. Correspondingly, for any data domain group in the map 1100 corresponding to a particular composite image grid position, the die presence 50 is indicated by a digit code or value, such as "1"; and the die vacancy 52 It is indicated by another digit code or value, such as "0". In this manner, for each lattice position of the enlarged image portion 1002 containing the remaining crystal grains 50 in the composite image 1000, the hollow wafer selection pattern 1100 represents the crystal grains 50 left at the lattice position by a digit "1". . Similarly, for each lattice position of the die vacancy 52 in the enlarged portion 1002 of the composite image 1000, the hollow wafer selection map 1100 represents the vacant die position 52 by a digit "0".

Referring again to FIG. 13, the fifth program portion 610 is directed to determining whether or not the grain location of each active region within the composite image 1000 has been subjected to grain presence 50 or grain vacancy 52. If no, the routine 600 returns to the first program portion 602.

Once the effective area die position is considered, the routine 600 includes a sixth program portion 620 that involves comparing the die presence and die vacancy designation of the wafer 1100 to the local PW map data for the corresponding die position. Thereby determining whether a grain removal error has occurred. For any particular effective area grain position, if there is an inconsistency between the presence or absence of a grain in the wafer selection pattern 1100 and the local PW pattern data indicating the presence or absence of a grain in the grain location, A grain removal error has occurred. The routine 600 can further include a seventh program portion 622 that involves generating and storing a die extraction/selection error map that identifies or indicates in code which of the hollow wafer die locations have undergone a die removal error. In various embodiments, the die selection error map provides a set of digital coded for the entire hollow wafer 10, the digital code indicating that: (a) the die 20 should be selected from the dicing wafer 5 but not properly left in Each grid position on the wafer 10 is hollowed out; and (b) each grid position of the die 50 that is left but vacant.

In some embodiments, the sixth program portion 620 is directed to comparing the local PW map data for the die presence/vacancy design and the corresponding die position for each grain location of the hollow wafer pick pattern. Based on the comparison result, the SWIS 200 can generate a grain removal error map. In other embodiments, the SWIS 200 may generate a predetermined selection map directly from the local PW map data itself, wherein the predetermined selection map indicates the predetermined grain presence 50 and the predetermined grain vacancy 52 based on the local PW map data, and is The same coding scheme used in Figure 1100 (i.e., the same die presence and die vacancy designation) is selected for the hollow wafer. In the above embodiment, the SWIS 200 simply subtracts the hollow wafer selection map 1100 from the predetermined selection map (or vice versa) to generate a die extraction error map, thereby determining whether a grain removal error has occurred. After subtraction, any non-zero result of any grain position in the die extraction error map indicates that a grain removal error has occurred at the die position.

Referring again to FIG. 9, the routine 300 can include an eleventh program portion 360 that involves transmitting a die fetch error data, such as a die fetch error map, to one or more other destinations or systems; and/or storing a die fetch error. The figure is in a location where another device or system can be accessed (for example, in a data file stored in a regional or network drive). Storing or transmitting die-selection error map data or die vacancy error flags facilitates grain classification error correction operations, for example, from tape or other destinations that are only intended to accommodate well-qualified and visually acceptable dies 26 The defective crystal grains 24 and/or the crystal grains 25 which are electrically good but are visually unacceptable are taken back.

According to an embodiment of the present invention, 100% of the hollow wafer 10 can be inspected in a very time-saving manner. For example, in a typical implementation, 100% of the 8 hollow wafers 10 with 1 x 1 mm2 grains are automatically inspected to identify voided wafer grain presence/grain gap errors, which can be in less than 3 minutes. Finished. In addition, according to an embodiment of the present invention, 100% of the hollow wafer 10 can be automatically inspected and the corresponding grain presence/grain gap error is identified without significantly affecting the manufacturing capacity (the lowest, negligible, for manufacturing capacity) Approaching zero, or substantially/actually zero impact).

Automated inspection of hollowed strips

In many semiconductor manufacturing situations, the component or device can be carried, supported, or supported by a substrate other than the wafer, and the SIS/SWIS 200 can perform a hollow inspection operation to verify that the component is properly removed from the substrate. For example, as shown in Figures 14A and 14B, element 1220 (e.g., bump die 20) can be carried, supported, or supported by resilient webs 1200a,b. Each strip 1200a,b includes a plurality of different strip segments 1202a-d that are separated by a notch 1204. Each strip segment 1202a-d provides a plurality of grid locations on which element 1220 can reside. The lattice locations within each strip segment 1202a-d are defined by horizontal and vertical grid lines 1230, 1232, respectively, in a manner that is immediately apparent to those of ordinary skill in the relevant art. There are different types of strips 1200a, b comprising: (a) strips 1200a having holes, openings, or reference marks 1206 formed at specific locations around or outside edges thereof; and (b) strips 1200b lacking the above Peripheral holes, openings, or reference signs 1206.

The one or more strips 1200a, b and the component 1220 that are carried, supported, or supported can be adhered to the film frame 12 in a manner substantially the same as or similar to the wafer, so that the adhesive film 11 supports the strip. Plates 1200a, b and their components 1220 are shown in Figures 14C and 14D. The slats can be cut along the transverse and longitudinal grids 1230, 1232 within each slat section 1202a-d. Cutting within the gap 1204 between each of the strip segments 1202a-d can be avoided. As if the dicing wafer 5 is adhered to the film frame 12, after cutting the slats, the cutting grooves 34 will extend along the lateral and longitudinal grids 1230, 1232 within each slat section 1202a-d to The depth of the adhesive film 11.

If the cutting strip 1205a has a reference hole 1206, the film frame 12 to which the cutting strip 1205a is adhered does not need to have indentations or other reference features because image processing is used to determine the cutting strip 1205a relative to the film frame 12. The orientation, as well as the orientation of the determining component 1220 relative to the film frame 12, can occur with reference apertures 1206 that utilize image processing to identify the slats. If the cutting strip 1205b lacks the reference hole 1206, the film frame 12 can include some indentations 13 and/or other reference features to utilize image processing to determine the orientation of the elements 1220 of each strip relative to the film frame 12. .

Each of the cutting strips 1205 has a corresponding processing pattern similar to that described above for the processing wafer, which illustrates the elements 1220 within each of the strip segments 1202a-d of the cutting strip 1205. Electrical test and visual inspection results. The processing map corresponding to the strip 1205 and the PW map corresponding to the wafer may be referred to as a processing map, a component processing diagram, a processing component diagram, or a processed die pattern, as will be immediately understood by those of ordinary skill in the relevant art. .

The active component area can be defined as the area in which the strip 1200 carries the component, and the component will be selectively removed from the area in a set of component sorting operations (eg, die sorting operations). Each strip segment 1202 or portion thereof can be defined as an active element region. The active component area corresponds to the component processing map in a manner substantially the same as or similar to that described above for the PW diagram. The first/starting element position of a particular strip segment 1202 is typically defined as a predetermined corner grid position (eg, an upper left grid position); however, the first/starting element position may also be defined as another grid position.

After the component sorting operation (eg, die sorting operation), that is, after the component 1220 is selectively removed or removed from the cutting strip 1205, the cutting strip 1205 has a "hollowed" appearance, which may be referred to as a hollowed out panel 1210, as shown in the figure. 14E is shown. In Figure 14E, the grid locations 1250 for vacancies or vacancies within each of the slab segments 1202a-d are indicated by blank or unshaded regions, while the elements 1252 present or left by the components are indicated by shaded regions.

In accordance with various embodiments of the present invention, an automated inspection of the hollow strip 1210 can be performed to verify that the component 1220 has been properly removed in a manner substantially the same as or similar to that described above for the hollow wafer 10. More specifically, after the component sorting operation is completed, the segmented image 1310 of the hollowed out panel 1210 can be taken into consideration, as shown in FIG. 14F. During the segmentation of the segmented images 1310a-d of the particular hollowed out panel 1210, the hollowed out panel 1210 is positioned relative to the field of view of the SWIS image capture device such that the segmented image 1310a captured by the SWIS image capture device 220 is d A portion or segment that spans the overall surface area of the hollowed strip, especially a portion or segment that spans the active element area of the hollowed strip. In the exemplary embodiment illustrated in Figure 14F, a total of four segmented images 1310a-d are taken, each corresponding to a spatial region that is slightly larger than the region of a single strip segment 1202a-d. One of ordinary skill in the relevant art will immediately appreciate that, in view of the above, in other embodiments, the strip geometry (i.e., length and/or width), SWIS image capture device field of view, and/or component/ Depending on the grain size, more or less segmented images 1310 can be taken. In this manner, in other embodiments, the individual segmented image 1310 can capture segments of the slab segments 1202a-d (eg, 1/2, 1/3, 1/4, 1/5, etc...), or shoot More than one strip segment 1202a-d.

During the segmentation of the segmented image of the hollow strip 1210, illumination can be directed from the film frame 12 and the hollow strip 1210 adhered thereto to the SWIS image capture device 220. Illumination can also be directed from above the film frame 12 to the hollow strip 1210 adhered to the film frame 12 such that illumination can be reflected from the exposed surface of the element 1250 to the SWIS image capture device 220.

The captured segmented images 1310a-d are stitched together to form a composite image in substantially the same or similar manner as described above; the hollow panel 1210 is first in each of the strip segments 1202a-d The starting component position (e.g., the lattice position corresponding to a predetermined corner, such as the upper left lattice position) can be determined within the composite image. The reference origin or grid position can also be determined before or at the same time as determining the first/starting element position. An automated hollow strip analysis operation can be performed for each strip segment 1202a-d. During this operation, the SWIS 200 uses image processing to analyze the pixel arrays within the composite image of the hollowed-strip, starting from the first/starting component positions of the strip segments 1202a-d, determining the strip segments 1202a-d in question. Whether each grid position within has indicated that the component has a 1250 or component vacancy 1252. The SWIS 200 may additionally generate a strip selection map for the strip segments 1202a-d or the hollow strips 1210 in a manner similar to that described above. Finally, after generating the strip selection map, the SWIS 200 can compare the strip selection map to the strip processing map and generate a selection error map corresponding to the hollowed out strip 1210 in question, in a manner similar to that described above. Selecting an error data set can be transferred to one or more other systems or devices for removal of the error correction work.

In some embodiments, the SWIS 200 is integrated in series with the component sorting device 115, and the SWIS 200 can generate a plurality of hollowed-strip composite images, each composite image corresponding to a predetermined strip segment 1202 of the hollowed out strip 1210 in question. (eg 1 or 2 strip segments 1202). For example, the first hollow strip composite image may correspond to the first strip segment 1202a; the second hollow strip panel composite image may correspond to the second strip segment 1202b; and so on. In the above embodiment, depending on the geometry of the field of view of the SWIS image capture device 220, each composite image may be a single segmented image 1310a-d. Once the component classification of the particular strip segment 1202a is completed, the SWIS image capture device 220 can capture a set of segmented images (eg, one segmented image or more than one segmented image) and generate a corresponding image The composite image of the hollowed-slab section 1202a can be immediately or subsequently subjected to component sorting operations of the other/next cutting strip section 1202b. As a parallel calculation program concurrent with the component sorting operation of the next cutting strip segment 1202b, the SWIS 200 can analyze the composite image corresponding to the hollowed-slab segment 1202a to determine the component with reference to the appropriate portion of the component processing map. Whether 1220 is correctly removed from the hollow strip section 1202a. In this manner, the SWIS 200 can determine the last segment of the stripping operation while the other strip segment 1202b-d is performing the component sorting operation, that is, before the other strip segment 1202b-d completes the component sorting operation. 1202a Whether an error has occurred in order to take corrective action more quickly.

In addition to the above, the SWIS 200 can also generate and analyze the composite corresponding to the hollow strip 1210a carried by the film frame 12 while performing the component sorting operation on the next cutting strip 1210b carried by the same film frame 12. Image (for example as a parallel computing program).

In some embodiments, in addition to generating one or more composite images corresponding to the hollowed out panel 1210, the SWIS 200 may also produce one or more composite images of the cutting strip 1205 prior to the start of the die removal operation ( For example, a composite image of a dicing strip 1205 that is completely or substantially completely present. The segmented images 1300a-d are taken to produce one or more composite images of the cut strip as shown in Figure 14G. The segmented images 1300a-d can be stitched together in a digital manner to form a composite image of the cut strip in a manner similar to that described above. As in the case of a composite image of a wafer, the composite image of the cut strip can be used to improve the accuracy of determining or identifying the reference origin and/or the position of the starting component of the composite image of the hollowed strip; and/or one or more Used as a navigation aid or tool in component classification related procedures.

Cutting wafer composite image as navigation aid/tool

The dicing of the wafer composite image essentially includes dicing the location of the die 20 of each active region on the wafer 5 and accurately encoding it, thus also exhibiting or staking each lattice location on the wafer 10 coding. Therefore, cutting the wafer composite image facilitates accurate identification and/or effective navigation (eg, direct navigation) to (a) hollowing the wafer composite image and/or (b) actually cutting the wafer 5 or actually stenking the wafer 10 Within the boundaries of any or each effective area grain location.

For example, when a die sorting operation is being performed on the dicing wafer 5, the die sorting device 115 can transfer the one or more dies 20 or the dicing wafer position of each die 20 as determined by the pick and place device to SWIS 200. The dicing wafer grain position may be presented as a relative coordinate (eg, row and column coordinates relative to the first/starting grain (0, 0) grid position) or a grain classification encoder position/real space corresponding to the die 20 coordinate. The SWIS 200 can then compare, track, monitor, or plot the image space grain position within the composite image of the particular diced wafer, or the calculated predetermined code corresponding to the grid position of each active region of the diced wafer composite image. The position of the device, corresponding to the die position 20 considered by the die sorting device 115, is for example based on immediate or near instantaneous time during the die sorting operation.

In some embodiments, the SWIS 200 can calculate an expected relative for any particular die or grid position within the composite image of the diced wafer based on the diced wafer composite image and the conversion factor of the predetermined image space to the actual space. The actual space (x, y) position or coordinate set, or a relative encoder position offset, corresponds to its corresponding die/grid position on the actual dicing wafer 5 relative to the first/start grain position. The conversion factor from image space to real space may depend on the resolution of the image capture device, while the predetermined relative encoder position offset may depend on the encoder specifications in a manner that is immediately apparent to one of ordinary skill in the relevant art. In some embodiments, prior to the start of the die sorting operation on the dicing wafer 5, the SWIS 200 can calculate a predetermined relative physical space (x, y) position or coordinate set, or a predetermined relative encoder position offset. Shift, corresponding to each die/grid position on the dicing wafer 5; and storing the calculation data in the relative encoder position map.

If during or after the die sorting operation, the die sorting device 115 needs to move from one of the current die positions on the dicing wafer 5 across one or more die or grid locations to a particular target die or grid location. The die sorting device 115 can transmit the target die/lattice position to the SWIS 200, and the SWIS 200 can determine or calculate an appropriate predetermined relative real space (x, y) position or a predetermined relative encoder position offset, corresponding to The target die/lattice position on the wafer 5 is actually cut and transmitted to the die sorting device 115. The die sorting device 115 may utilize this predetermined relative real space (x, y) position or a predetermined relative encoder position offset to generate a set of updated encoder positions. Thus, the extension station 210 and the die sorting device 115 can move relative to one another in accordance with the updated set of encoder positions such that the die sorting device 115 can navigate directly into the edge of the target die/lattice position (eg, relative to Target die/grid position size, approximately +/- 10% - 30% or +/- 20% of the center of the target die/lattice position). In this manner, in accordance with some embodiments of the present invention, the die sorting device 115 does not require the next closest die as a reference to navigate across multiple grid locations, which can save significant time, thereby increasing throughput and significantly reducing indexing or The possibility of location translation errors. After direct navigation, the particular location of the die 20 remaining at the target die/grid position of the dicing wafer 5 can be verified via the die sorting device imaging system to accurately establish the position of the pick and place device relative to the die 20. .

One of ordinary skill in the relevant art will appreciate that (a) whether the extension station 210 is moving/movable while the pick-and-place device remains stationary, or (b) whether the pick-and-place device is moving during the extension station 210 remains stationary / Moveable, the above encoder position or updated encoder position may correspond to or be applied to the movement of the extension station 210 or the movement of the pick and place device.

In a batch of dicing wafers 5 that receive the same process together, the composite image of the diced wafer of the first dicing wafer 5 can be defined as a composite image of the batch navigation dicing wafer, in the first or kth piece of the batch During the grain sorting operation of the dicing wafer (where k > 1), it is used to assist or guide direct navigation from one die/grid position to another die/grid position. In some embodiments, the SWIS 200 can generate and store a predetermined relative encoder position offset for each die/grid position within the active area of the dicing wafer, based on the batch navigation composite image, wherein the relative encoding The positional offset is referenced to the first/starting die.

If the batch navigation composite image is generated according to the dicing wafer 5 that is undergoing the die sorting operation, the rotation orientation of the batch navigation composite image is the same as the dicing wafer 5 on the expansion stage 210 or from the dicing wafer. The hollow wafer 10 produced by the circle 5 has the same rotational orientation. That is, the rotational orientation difference between the batch navigation composite image and the dicing wafer 5 on the expansion stage is zero. However, if the batch navigation composite image is generated according to the specific dicing wafer 5, which is different from the dicing wafer 5 on the extension stage 210 that is accepting or has completed the singulation operation, the batch navigation composite image and the extension There may be a rotational azimuth difference or offset between the dicing wafer 5 or the hollow wafer 10 being carried by the stage 210. In order to accurately navigate directly from the current or specific die/grid position to the next or target die/grid position, the rotational offset θ can be determined. For example, if the batch navigation composite image is generated according to the first dicing wafer 5 in a batch, and the dicing wafer 5 that is undergoing the die sorting operation is the kth dicing wafer 5 in the same batch The SWIS 200 can generate a composite image for the kth sliced wafer; determine a center point of the kth sliced wafer in its composite image; and register via one or more conventional image registration techniques The batch navigation composite image and the kth slice are cut into a composite image to determine a rotational offset θ between the composite image of the kth sliced wafer 5 and the batch navigation composite image. Therefore, the composite image of the kth slice dicing wafer 5 may be referred to as a rotation check composite image. Once the rotational offset θ is known, the extension stage 210 can be rotated to correct or compensate for the rotational offset θ, and the die sorting device can be cut from the kth slice by one of the methods described above. The specific die/lattice position navigates directly to another die/lattice position.

The batch navigation composite image can also serve as an aid or guide for direct navigation of the grain sorting device on the hollow wafer 10 in a manner similar to that described above.

Typical inspection of the grain classification equipment to detect the correct reference grain

In addition to the above, the batch navigation composite image can be used to assist in checking whether the die sorting device 115 has been detected or not before performing the die sorting operation to selectively extract the die 20 from the kth sliced wafer 5. The correct or actual reference die 21 on the kth sliced wafer is selected. More specifically, a diced wafer composite image can be produced for the kth sliced wafer, and a corresponding reference origin (eg, as described above) can be automatically detected, such as dicing a wafer in the composite image. The center point in it. The batch navigation composite image and the kth sliced wafer composite image may be registered with respect to each other, and the extension stage 210 may be rotated to correct or compensate the batch navigation composite image and the kth sliced wafer composite Any rotation offset between the images θ.

In the die classification/composite image generation/composite image navigation/hollow wafer inspection configuration or set prescription, a relative encoder position offset may be included, which corresponds to the remaining or predetermined remaining on the reference dicing wafer 5. A plurality of auxiliary reference dies such that the (x, y) coordinates or relative encoder positions of each of the auxiliary reference dies are offset relative to the first/start dies and/or reference on the dicing wafer 5 The position of the die 21 is accurately defined, measured, or known. The auxiliary reference die may include a particular type of die having features that facilitate automatic detection by the die sorting device 115, such as mirrored die or empty die.

The die sorting device 115 can detect the reference die 21 on the kth slice dicing wafer, for example, by conventional means, the reference die may or may not be the correct, true, or actual reference crystal of the kth slice dicing wafer. Granule 21. Therefore, the detected reference die 21 can be defined as the candidate reference die 21. After detecting the candidate reference die 21, the die sorting device 115 can use the relative encoder position offset corresponding to the particular auxiliary reference die to navigate directly from the position of the candidate reference die 21 to some predetermined aid. The position of the cut wafer with reference to the die position. The subsequent die sorting device imaging system can capture and analyze an image of the predetermined auxiliary reference die position, and can capture and analyze some images at the actual lattice position offset adjacent to the predetermined auxiliary reference die position, A determination is made as to whether the die 20 at a predetermined auxiliary reference die location is a substantially predetermined auxiliary reference die. If the die 20 at the predetermined auxiliary reference die location is not a predetermined auxiliary reference die, the candidate reference die 21 is not or is likely to be a correct or actual reference die. The die sorting device 115 may repeat the above procedure to navigate to predetermined positions of the plurality of auxiliary reference dies to improve the accuracy and reliability of determining whether the candidate reference dies 21 are the correct reference dies 21. If the die sorting device 115 confirms that the candidate reference die 21 is not actually the true reference die 21, then another candidate reference die 21 (eg, automatic and/or manual) may be detected or selected and the above procedure may be repeated.

If for each of the auxiliary reference dies, the auxiliary reference dies have been accurately detected or identified at the corresponding predetermined auxiliary reference dies, the candidate reference dies 21 are or are likely to be the kth diced wafer Actual reference to the die 21. Once it is confirmed that each of the auxiliary reference dies has been accurately detected after navigating to its predetermined position, the selective removal of the dies 20 from the kth dicing wafer 5 can begin.

Therefore, according to the embodiment of the present invention, the possibility of detecting or selecting the incorrect reference die 21 can be reduced or greatly reduced before the die removal operation, so that a serious systematic error in the take-out operation can be reduced or greatly reduced. possibility.

One of ordinary skill in the relevant art will recognize that after navigating from the candidate reference die 21 position of the kth sliced wafer to the predetermined position of the first/starting die of the kth sliced wafer, navigating to one or more predetermined The act of assisting the reference die position may begin at a different position on the kth slice dicing wafer, such as the first/start grain position of the kth dicing wafer.

Typical hollow wafer restoration operations

After completing the die sorting operation of the kth dicing wafer 5, the kth dicing wafer 5 becomes the kth stencil wafer 10, which typically reduces many, most, or nearly/substantially all of the effective area die 20 . If the SWIS 200 determines that the k-th hollow wafer 10 has undergone a grain removal error (for example, a predetermined number or percentage of die removal errors) based on the k-th hollow wafer composite image analysis, the SWIS 200 and the pick-and-place device The batch navigation composite image can be used as a navigation guide to conduct or direct the wafer 10 recovery operation, and the die 20 that was previously selected and sent to a particular destination, such as a tape, is placed back. The recovery operation of the kth stencil wafer 10 is basically to recreate the kth dicing wafer 5, or to create a substitute kth dicing wafer 5, so that the die 20 returning to the film frame 12 is effective The spatial arrangement of the regional grain locations has a high degree of positional accuracy. Thus, the alternate dicing wafer 5 can re-accept the die sorting operation to remedy one or more die removal errors.

More specifically, as a part of the winding operation, that is, the die 20 is removed from the k-th dicing wafer 5 and sent to a tape via a take-up assembly 120, the die ID and corresponding can be stored. The dicing wafer position/grid position of each die 20 (eg, stored in a memory or database in the form of a fetched die data file). The actual dicing wafer position/grid position of each die 20 can be defined by the encoder position of the die 20. For each die 20 on the k-th slice wafer 5 that has been fed to the tape, the restoration of the kth wafer 10 may involve retrieving the die 20 from the tape and retrieving the stored corresponding crystal The grain ID and the dicing wafer position/grid position, and the dies 20 are placed back into the appropriate dicing wafer position/grid position on the kth stencil wafer 10 via the pick and place apparatus 160.

If the kth hollow wafer 10 is not removed after the completion of the first group of die sorting operations, and is still firmly retained by the extension 210, it can be known that the kth hollow wafer reference origin/lattice position is defined. The encoder position, such as the center point of the kth hollow wafer. Therefore, the kth stencil wafer 10 can be left on the extension stage 210 and directly restored.

If the restoration operation of the kth hollow wafer 10 is performed after the kth hollow wafer is removed from the extension 210, and the kth hollow wafer is sent to the extension 210 or another extension before the restoration The SWIS 200 can generate a composite image of the kth hollow wafer 10 and determine a reference origin of the kth hollow wafer within the composite image, such as its center point. The k-th hollow wafer composite image can be registered with the batch navigation composite image via one or more conventional image registration techniques, and the position of the reference die 21 on the kth hollow wafer 10 can be checked or confirmed. The manner is basically the same as or similar to that described above.

After checking the position of the reference die 21 on the kth hollow wafer 10, the die 20 can be placed back onto the film frame 20 according to its stored encoder position, so that the restored die 20 is accurately located in its active area. Plaid position.

Typical SWIS Attachment Inspection System Example

In some embodiments, the SWIS 200 can incorporate or integrate any type of inspection system or device that is used to process and inspect the components carried by the film frame 20, such as the die 20, even if the inspection system or The same is true for the expansion station 210 not included in the device. Thus, the inspection system can be used not only to perform originally scheduled inspection operations (e.g., to capture images to identify micron or sub-micron scales on the die 20), but also to perform independent hollow wafer inspection operations, with the die sorting device 114. Separation.

In the above embodiments, the inspection system includes or is modified to include: (a) an image capture device 220 for or for capturing a segmented image of the hollow wafer 10; and (b) processing resources for Or may be used to (i) generate a composite image corresponding to the hollowed wafer, (ii) determine a reference origin and/or a first/starting die position on each of the hollowed wafers 10; (iii) receive or Retrieving PW map data corresponding to the hollow wafer 10, and (iv) comparing the data content of each of the hollow wafer composite images with the data content of the corresponding PW map to identify whether one or more crystal grain removal errors have occurred, The manner is basically the same as or similar to that described above.

In order to assist in understanding, in a typical example, an optical inspection system as described in International Patent Publication No. WO/2010082901 is provided for carrying, stabilizing, and accurately positioning the wafer frame 12 via the wafer table assembly. Wafer and dicing wafer 5. The wafer table assembly can include a polar or hyperplanar wafer table as described in International Patent Publication No. WO/2014035346. The wafer stage assembly can carry, securely retain, and accurately position the hollow wafer 10 in a manner that is immediately apparent to those of ordinary skill in the relevant art.

When the hollowed wafer 10 is carried by the wafer table and is firmly retained, the film 11 is not expanded as the dicing wafer 5 that caused the hollow wafer 10 during the die sorting operation remains on the extension stage 210. The film 11 is in a slightly or partially relaxed state. Therefore, the inter-die separation distance on the hollow wafer 10 will be different from the hollow wafer 10 carried by the extension stage 210. Even so, the grid lines 30, 32 and the relative row and column positions on the hollow wafer 10 remain unchanged, substantially constant, or substantially unchanged. Thus, the wafer table can be placed under the SWIS image capture device 220 that has been incorporated into the inspection system to take a segmented image of the hollow wafer 10. A composite image of the hollowed wafer is then generated and the position of the reference origin and the first/starting grain within the composite image of the hollowed wafer is determined. Then, the contents of the hollow wafer data corresponding to the grain position of each effective region (for example, the pixel array value) and the data content of each corresponding grain position in the hollow wafer PW map can be aligned to determine the grain sorting operation period. Whether one or more crystal removal errors have occurred.

Typical composite image generation process

15A and 15B are flow diagrams of a process 1400, 1500 for generating a composite image in accordance with an embodiment of the present invention. The procedures 1400, 1500 relate to: (a) a batch of grain sorting operations in which the batch navigation composite image is used as an aid or guide during the pick and place operation; (b) each of the hollow crystals generated for the grain sorting operation Circle 10 performs a hollow wafer inspection operation; and may involve (c) film frame restoration operations.

In one embodiment, the first program portion 1402 involves retrieving composite image generation/navigation parameters from a prescription and hollowing out wafer inspection parameters. The second program portion 1404 involves selecting the first or next slice of the cut wafer 5 in the batch and sending it to the extension station, and the third program portion 1410 involves generating a batch navigation composite image. The batch navigation composite image is produced in the same, substantially identical, or similar manner as described above, as shown in Figure 16 in accordance with a composite image generation program 1600 in accordance with an embodiment of the present invention.

The fourth program portion 1420 involves performing a die sorting operation for the current dicing wafer 5, using the batch navigation composite image as an aid or guide, directing the die sorting device 115 directly to each of the predetermined detachment from the film frame 20. The position of the die 20 is handled in accordance with the PW map corresponding to the dicing wafer 5. After completing the die sorting operation, the fifth program portion 1430 involves generating a composite image of the hollowed wafer in a manner as shown in FIG.

The sixth program portion 1440 involves examining the effective area die position of the 100% hollow wafer. The 100% open wafer inspection is performed in the same, substantially the same, or similar manner as described above, as shown in Figure 17 in accordance with an open wafer inspection program 1700 in accordance with an embodiment of the present invention. It should be noted that with regard to analyzing the grain classification error, such as the second program portion 1720 of FIG. 17, the relative encoder position map helps to determine the inconsistency between the composite image data content of the hollowed wafer and the data content of the PW image. nature. For each grain presence 52 or die vacancy 50 shown in the hollow wafer pick-up, the relative encoder position of the die/grid position in the relative encoder position map can be compared to the corresponding grain stored in the PW map. The classification device encoder positions are compared. For example, for the die 50 to be left as shown in the hollow wafer pick-up, the difference between the relative encoder positions in the encoder position map for the die 50 should be left to correspond to the PW map. The grain/grid position of the grain classification device is compared between the encoder positions. For the hollow wafer picking diagram, the relative encoder grain position of the die 50 is determined to be greater than a certain predetermined amount of the encoder position difference, such as a spatial positioning error. Exceeding about 20% of the grain size, this means that after the die sorting operation is completed, a certain defective die 20 is improperly taken out from the film frame 12 instead of remaining on the film frame 12.

Referring again to Figure 15A, depending on the number and/or nature of the die selection errors identified by the hollow wafer inspection, the hollow wafer 10 can be instructed or approved for recovery operations. The restoration operation can be performed via the seventh and eighth program portions 1450, 1460. In the ninth program portion 1462, the restored film frame 12 can be re-evaluated for subsequent grain sorting operations (e.g., immediately after recovery or later). In some embodiments, a certain number of die selection errors have occurred if the restore operation is approved or indicated. This may indicate one or more problems with the previously generated batch navigation composite image. Accordingly, another piece of the dicing wafer 5 can be selected via the ninth program portion 1462 to generate a substitute lot navigation composite image and returned to the second program portion 1404.

If, in the seventh program portion 1450, the film frame restoration operation is not approved or instructed, in the tenth to twenty-first program portions 1502 to 1570, the next piece of the dicing wafer 5 may be selected; and the rotation check composite pattern is generated. Detecting whether the correct reference die 21 has been detected on the selected dicing wafer 5; the die classification of the selected dicing wafer 5; the hollowing check corresponding to the stencil wafer 10; and possible recovery operations, As shown in Figure 15B. The tenth to twenty-first program portions 1502 - 1570 can be repeated for each additional wafer 5 in the batch.

18 is a flow diagram of a simplified routine 1800 for verifying that a correct reference die 21 has been selected on a dicing wafer 5 in accordance with an embodiment of the present invention. The program 1800 includes first through eleventh program portions 1802 - 1840 that relate to determining whether the candidate reference die is the actual reference die 21 in substantially the same or similar manner as described above.

19 is a flow diagram of a process 1900 for restoring a film frame 12 that has selectively removed the die 20 in accordance with an embodiment of the present invention. The program 1900 includes first through sixth program portions 1902 - 1920 relating to film frame restoration in a manner substantially the same as or similar to that described above.

Particular embodiments of the present invention address at least one aspect, problem, limitation, and/or disadvantage of existing hollow wafer inspection techniques. Although the present invention has been described in terms of its features, aspects, and/or advantages, other embodiments may present such features, aspects, and/or advantages, and not all embodiments are required to present the features, aspects, and / or advantages are considered to fall within the scope of the present invention. One of ordinary skill in the relevant art will appreciate that many of the systems, components, processes, or alternatives disclosed above can be combined with other different systems, components, processes, and/or applications. In addition, various changes, modifications, and/or improvements may be made to the various embodiments of the invention disclosed herein. The various systems, components, processes and/or variations, modifications, and/or improvements are included in the scope of the invention and the following claims.

10‧‧‧ hollow wafer
100, 101‧‧‧Previous process semiconductor component manufacturing system
1000‧‧‧Composite image
1002‧‧‧Magnification
102‧‧‧ first visual inspection
104‧‧‧Second visual inspection, partial/first cutting
106‧‧‧Parts/first cut, second visual inspection
108‧‧‧Electrical test
11‧‧‧film
110‧‧‧complete/second cut
1100‧‧‧ hollow wafer selection map
112‧‧‧Additional/Third Visual Inspection, Final/Third Visual Inspection
114, 114a, 114b, 114c‧‧‧ component sorter / die classification
115‧‧‧Graph Classification and Inspection System
12‧‧‧film frame
120‧‧‧ Tape and Reel Components
1200a, 1200b‧‧‧ boards
1202, 1202a, 1202b, 1202c, 1202d‧‧‧ sections
1204‧‧‧ gap
1205a, 1205b‧‧‧ cutting strip
1206‧‧‧Reference hole
1210a‧‧‧ hollow board
122‧‧‧Visually unqualified disc/box
1220‧‧‧ components
1230‧‧‧ transverse grid
1232‧‧‧ vertical grid
1250‧‧‧ vacant grid position
1252‧‧‧ Elements that exist or remain
13‧‧‧ dent
Segmentation images of 1300a, 1300b, 1300c, 1300d‧‧‧
Segment images of 1310a, 1310b, 1310c, 1310d‧‧
1400‧‧‧Programs for generating composite images
1402-1462‧‧‧Steps
150‧‧‧film frame box
1500‧‧‧Programs for generating composite images
1502-1570‧‧‧Steps
160‧‧‧ pick and place equipment
1600‧‧‧Composite image generation program
1610-1630‧‧ steps
170‧‧‧ first camera
1700‧‧‧ hollow wafer inspection program
1710-1720‧‧ steps
172‧‧‧second camera
174‧‧‧ third camera
18‧‧‧Small Wafer Zone
180‧‧‧ fourth camera
1800‧‧‧Simplified procedure
1802-1840‧‧‧Steps
182‧‧‧ fifth camera
1900‧‧‧Program
1902-1920‧‧‧Steps
20, 24, 25, 26, 50‧‧ ‧ grains
200‧‧‧ hollow wafer inspection
21‧‧‧Reference grain
210‧‧‧ Extension station
212‧‧‧ surrounding support members
214‧‧‧ architecture
215‧‧‧ openings
220‧‧‧Image capture device
230, 232‧‧‧ illumination source
250‧‧‧Processing unit
262‧‧‧ hollow wafer inspection module
264‧‧‧Check configuration memory
266‧‧‧ image memory
268‧‧‧ working memory
270‧‧‧Transport/Network Interface Unit
30, 32‧‧ ‧ grid
300‧‧‧Automation procedures
302-360‧‧‧Steps
34‧‧‧Cutting trench
40‧‧‧ In-grain features
400‧‧‧ First procedure
402-438‧‧‧Steps
410‧‧‧ Second procedure
5‧‧‧Cutting wafer
500‧‧‧Composite image generation program
502-514‧‧‧Steps
52‧‧‧ vacancy location
600‧‧‧Program
602-622‧‧‧Steps
900a, 900b, 900c, 900d, 900e, 900f, 900g, 900h, 900i‧‧‧ segmentation images

1 shows a block diagram of a partial process of a prior art semiconductor component fabrication system involving a particular aspect of the backside semiconductor process, including die classification. Figure 2A shows a typical dicing wafer on which several components or dies have been fabricated that are bonded to a film frame. Figure 2B shows a typical hollow wafer corresponding to the dicing wafer of Figure 2A. Figure 3 shows a column of grains, each of which includes a set of features that can be interpreted as grain edges by machine vision algorithms. Figure 4 shows a hollowed-out wafer bonded to a film frame with Z = 5 typical hollowed-out wafer areas for comparison with Z = 5 corresponding processing areas. 5A is a block diagram showing a portion of a typical semiconductor component fabrication system including an automated open wafer inspection system (SWIS) in accordance with an embodiment of the present invention. Figure 5B is a generalized block diagram of a die sorting and inspection system for use in tandem hollow wafer inspection in accordance with an embodiment of the present invention. 5C is a block diagram of a portion of a typical semiconductor fabrication system associated with die classification, including a plurality of die sorting devices for delivering a hollow wafer to a common hollow wafer inspection system (SWIS) in accordance with an embodiment of the present invention. ). Figure 5D is a schematic illustration of a typical open wafer inspection system (SWIS) component in accordance with an embodiment of the present invention. Figures 6A and 6B show typical partial depth channels formed in the film of the film frame due to the cutting process, cutting the grooves. 7A shows a composite image of a hollowed out wafer, under typical illumination conditions, corresponding to providing illumination to the top and bottom ends of the hollowed out wafer to capture a segmented image, in accordance with an embodiment of the present invention. Figure 7B shows an enlarged portion of the composite image of the hollowed wafer of Figure 7A, corresponding to a 2 x 7 array of hollowed-wafer grid locations, defined by lateral and longitudinal cut grooves or grid lines. Figure 7C shows a typical digital encoding, for the enlarged portion of the composite image of Figure 7B, showing the presence of grains and grain vacancies in a typical partially hollowed wafer selection map. Figures 8A and 8B show portions of a typical actual wafer standard file. 9 is a flow diagram of an automated process for automated hollow wafer inspection in accordance with an embodiment of the present invention. 10A and 10B are flow diagrams showing a typical process for verifying the soundness of a partial PW map in accordance with an embodiment of the present invention. 11A and 11B are schematic illustrations of exemplary methods of capturing a hollow wafer segmented image to produce a composite image in accordance with an embodiment of the present invention. Figure 12 is a flow diagram of a typical process for generating a composite image from a set of segmented images in accordance with an embodiment of the present invention. Figure 13 is a flow diagram showing a typical procedure for generating a hollow wafer pick pattern and identifying a die removal error in accordance with an embodiment of the present invention. Figures 14A - 14G show a typical slat element and a hollow strip carried by a film frame that automatically inspects component removal errors in accordance with embodiments of the present invention. 15A-15B are flow diagrams of a typical process for producing a composite image in accordance with an embodiment of the present invention. 16 and 17 are flow diagrams showing a typical process for producing a composite image and inspecting a hollow wafer, respectively, in accordance with an embodiment of the present invention. 18 is a flow diagram of a typical process for automatically checking whether a reference die detected or selected by a die classification device is a correct reference die in accordance with an embodiment of the present invention. Figure 19 is a flow diagram of a typical process for restoring a film frame in accordance with an embodiment of the present invention.

300‧‧‧Automation procedures

302-360‧‧‧Steps

Claims (29)

An automated method for producing at least one composite image corresponding to at least one film frame carrying a cutting element, the method comprising: providing a film frame on which the cutting elements are arranged in a plurality of lattice positions, a plurality of grid positions including a set of effective area grid positions, the components are located inside; a set of segmented images of the film frame, each segment image corresponding to a predetermined portion of the space region spanned by the film frame And comprising at least a set of effective area grid positions; generating a composite image corresponding to the film frame from the segment image group, the composite image comprising image data indicating that (a) each of the dwellable elements is valid Area grid position, and (b) element presence or component vacancy for each effective area grid position; and processing the composite image via image processing techniques to determine a reference origin and a first effective area within the composite image a grid position in which the reference origin includes a predetermined point in the composite image or a predetermined effective area grid position, which represents (a) a film frame a corresponding predetermined point or predetermined effective area grid position, and (b) a predetermined effective area grid position corresponding to the processed image of the fabricated component, wherein the processed image includes a data structure, the storage corresponding to each The data content of the effective area grid position indicates that the component on which the valid area is located is qualified or unqualified for each valid area grid position. The method of claim 1, wherein the component is a semiconductor die corresponding to the wafer, and wherein the film frame carries at least (a) a portion of the wafer in the effective area of the wafer position group, or (b) The strip in which the component is located in the effective area grid position group. The method of claim 1 or 2, wherein the providing the film frame comprises at least one of the following: (a) providing a film frame before performing a group component sorting operation and selectively removing the component from the film frame. And (b) providing a film frame after performing the component sorting operation group. The method of claim 3, wherein the reference origin is different from the reference grain position used by the actuator sorting operation group on the film frame. The method of claim 4, wherein processing the composite image comprises: identifying at least a portion of the plurality of ruled lines in the composite image, corresponding to a plurality of cutting grooves between the elements on the film frame At least part of; defining a plurality of lattice positions according to a pixel region in the composite image, the boundary of which is surrounded by adjacent horizontal grid lines and adjacent vertical grid lines within the plurality of grid lines; and using a reference origin And determining a valid area grid position within the plurality of grid positions by a predetermined spatial relationship between the reference origin and the first effective area grid position. The method of any one of claims 1 to 5 wherein the segmented image group comprises directing illumination toward at least one of: (a) a top surface of the film frame, with the first set of illumination parameters as The reference, and (b) the bottom end surface of the first film frame, with the second set of illumination parameters as a reference. The method of claim 6, wherein the first set of illumination parameters and the second set of illumination parameters facilitate reliable image processing, discerning that the component is present at the grid position, the component is vacant from the lattice position, and each is defined The grid line of the grid position. The method of any one of claims 1 to 7, wherein the photographing the segment image group comprises using an image capturing device for photographing each segmented image as compared to micro or sub-microns for inspecting individual components. The image capturing devices used in the class have lower resolution and a larger field of view. The method of any one of clauses 1 to 8, wherein the segmented image group comprises a plurality of images corresponding to predetermined portions of the entire area of the film frame, and wherein the generating of the composite image comprises The digital mode splicing individual segmented images within a segmented image group. The method of claim 9, wherein the splicing of the individual segment images in the segmented image group in a digital manner comprises digitally connecting the individual segment images corresponding to at least partially adjacent regions of the film frame. By aligning the following individual segmented images: (a) overlapping along a common boundary group, corresponding to component edges and/or component features of the same component, and/or (b) directly adjacent to each other, with encoder values as A fiducial that indicates the relative actual spatial coordinates of the film frame during the capture of the segmented image set. The method of any one of claims 1 to 10, further comprising taking a prescription prior to producing the composite image, the prescription comprising at least one of the following: (a) grid data including some lateral grid lines and a plurality of vertical grid lines defining a plurality of grid positions; (b) image capture device parameters including image resolution of the image capture device relative to the component size and wafer size and a field of view of the image capture device; (c) at least a set of illumination parameters used to control the illumination characteristics incident on the film frame during the segmentation of the image group; (d) some segmented images to be captured; (e) individual segments within the segmented image group The degree of overlap between the segment images; (f) a set of reference origin parameters that indicate or define a reference origin position relative to a plurality of lattice positions; (g) a first effective region lattice position indicating effective a first component location within the regional grid location group; and (h) a set of parameters that can be used to examine the reference grain location detected by the die classification device. The method of any one of claims 3 to 11, further comprising: analyzing pixel data in the composite image of each effective region lattice position to determine a corresponding effective region lattice of the component present on the film frame The position or component is vacant; and the data content of the composite image is compared with the data content of the processed image for each valid region lattice position in the effective region lattice location group. The method of claim 12, wherein the comparing the content of the data comprises comparing a pixel value corresponding to a lattice position of each effective region in the composite image with a digit code corresponding to the lattice position of the effective region on the processing map. The method of claim 12 or 13, further comprising automatically determining: (a) whether one or more components are incorrectly removed from the film frame during component sorting, and/or (b) component sorting operations Whether one or more components are subsequently incorrectly left on the film frame. The method of any one of claims 3 to 14, wherein providing the film frame comprises providing a film frame prior to performing the component sorting operation group, wherein the generating the composite image is included before the actuator sorting operation group A composite image of the film frame is produced. The method of any one of claims 3 to 15, wherein the providing a film frame comprises providing a film frame after the actuator classifying operation group, wherein the generating the composite image is included in the actuator classifying operation group A composite image of the film frame is produced. The method of claim 15, further comprising using a composite image generated for the film frame prior to performing the component classification work group as a navigation aid or guide during a set of pick and place operations to locate at least one of the following : (a) an effective area target lattice position relative to the pick-and-place equipment during or after performing the component sorting operation on the same film frame or different film frames; (b) in the same film frame or different film frames During each of the upper component sorting operation groups, relative to each valid area grid position of the pick and place apparatus; and (c) during at least one set of film frame restoration operations, at least some effective areas of the plurality of grid positions relative to the pick and place apparatus The grid position, the restoration object is an element that is removed from the same or different film frames due to the component classification work group. The method of claim 17, wherein the film frame is a first film frame carrying a first dicing wafer in a plurality of dicing wafers, wherein the composite image of the film frame comprises the first dicing crystal A composite image of a circle, and wherein the composite image is used as a navigation aid or guide includes: storing a composite image of the first cut wafer as a batch navigation composite image, which can be used as a wafer for each wafer in the batch Navigation aid or guide; selecting a second film frame carrying the second cut wafer in the batch; generating a composite image of the second cut wafer; using image registration techniques to determine and correct the batch navigation composite image A rotational offset between the image and the composite image of the second diced wafer. The method of claim 18, further comprising: identifying a current die position on the second dicing wafer and a corresponding current grain position in the composite image of the second dicing wafer; identifying the second dicing crystal The target grain position of the circle and the corresponding target grain position in the batch navigation composite image; using the batch navigation image and the image space to the actual space conversion coefficient, the relative encoder position corresponding to the current die position is calculated And a relative encoder position corresponding to the target die position; generating an updated encoder position based on each calculated relative encoder position; and directing the pick and place device from the current die position using the updated encoder position Within a set of boundaries to the target grain location. The method of claim 18, further comprising checking whether the die sorting device detects the correct reference die on the second dicing wafer by: retrieving a relative encoder corresponding to the auxiliary reference dies from the prescription Position offset; directing the die sorting device from the candidate reference die position or the candidate reference origin to a predetermined position of the auxiliary reference die; automatically determining whether the auxiliary reference die appears at a predetermined position of the auxiliary reference die; Whether the auxiliary reference die appears at a predetermined position of the auxiliary reference die, and whether the candidate reference die is the correct reference die is checked. The method of claim 18, further comprising: selectively removing the die from the second dicing wafer to form a hollow wafer; generating a composite image of the hollowed out wafer; and utilizing image registration technology to Determine and correct the rotational offset between the composite image of the batch navigation and the composite image of the hollowed out wafer. The method of claim 21, further comprising: retrieving a relative encoder position offset corresponding to the auxiliary reference die from the prescription; picking and dropping the device from the candidate reference die position or candidate reference original on the hollowed out wafer Point directly to the predetermined position of the auxiliary reference die; automatically determine whether the auxiliary reference die appears at a predetermined position of the auxiliary reference die; and check the candidate reference die according to whether the auxiliary reference die appears at a predetermined position of the auxiliary reference die Whether it is the correct reference die; and starting a set of film frame restoration operations based on whether the candidate reference die is the correct reference die, the film frame recovery operation group includes removing the second wafer from the wafer The die is placed back into the second film frame to form a replacement second dicing wafer, and the hollowed-out wafer composite image and the batch navigation composite image are used as a film frame recovery operation group during navigation to Auxiliary or guide to different grain positions on the second film frame. The method of claim 3, wherein the generation of the composite image occurs simultaneously with the execution of the component classification work group via the parallel computing program. a system for producing at least one composite image corresponding to at least one film frame carrying a cutting element, the cutting element being located in an effective lattice area of a plurality of lattice positions, the system comprising: Processing unit; an extension station or wafer table for carrying and fixing the film frame; a set of illumination sources for guiding the illumination toward the film frame during the loading of the film frame by the extension table or the wafer table; An image capture device; and a memory that stores at least one set of program instructions that, when executed, causes the processing unit to perform the method of any one of claims 1 to 23. The system of claim 24, wherein the processing unit is coupled to a component sorting device for performing a group of component sorting operations on the film frame, the component sorting operation group relating to processing material content according to The film frame selectively removes the component. The system of claim 24 or 25, wherein the group of illumination sources comprises at least one of: (a) a first group of illumination sources for use in accordance with the first illumination parameter set during the segmentation of the segment image group; And guiding the illumination toward the top surface of the film frame, and (b) the second illumination source group for guiding the illumination toward the bottom end of the first film frame according to the second illumination parameter set during the shooting of the segment image group surface. The system of any one of claims 24 to 26, wherein the image capturing device has lower resolution and comparison than an image capturing device used to inspect micro- or sub-micron 瑕疵 of individual components. Large field of view. The system of any one of claims 24 to 37, wherein the memory stores a prescription comprising at least the following items: (a) grid data including some lateral grid lines and some longitudinal grid lines , which defines a plurality of grid positions; (b) image capture device parameters, including image capture device resolution relative to component size and wafer size, and image capture device field of view; (c) at least one set of illumination parameters Used to control the illumination characteristics incident on the film frame during the shooting of the segment image group; (d) some segmented images to be captured; (e) individual segmented images within the segmented image group (f) a set of reference origin parameters indicating or defining a reference origin position relative to a plurality of lattice positions; (g) a first effective region lattice position indicating a valid region lattice position group a first component location; and (h) a set of parameters that can be used to examine the reference grain location detected by the die classification device. The system of any one of claims 25 to 28, wherein (a) the system is integrated into a component sorting device for performing a component sorting operation group, or (b) the system is separated from the component sorting device.